maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/lib/Sema/SemaInit.cpp

4465 lines
173 KiB
C++
Raw Normal View History

Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
//===--- SemaInit.cpp - Semantic Analysis for Initializers ----------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
// This file implements semantic analysis for initializers. The main entry
// point is Sema::CheckInitList(), but all of the work is performed
// within the InitListChecker class.
//
move InitListChecker to be private to SemaInit.cpp llvm-svn: 65398
2009-02-25 06:48:58 +08:00
// This file also implements Sema::CheckInitializerTypes.
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
//
//===----------------------------------------------------------------------===//
Another step in the process of making the parser depend on Sema: - move DeclSpec &c into the Sema library - move ParseAST into the Parse library Reflect this change in a thousand different includes. Reflect this change in the link orders. llvm-svn: 111667
2010-08-21 02:27:03 +08:00
#include "clang/Sema/Designator.h"
Move Sema's headers into include/clang/Sema, renaming a few along the way. llvm-svn: 110945
2010-08-13 04:07:10 +08:00
#include "clang/Sema/Initialization.h"
#include "clang/Sema/Lookup.h"
Split out a header to hold APIs meant for the Sema implementation from Sema.h. Clients of Sema don't need to know (for example) the list of diagnostics we support. llvm-svn: 112093
2010-08-26 06:03:47 +08:00
#include "clang/Sema/SemaInternal.h"
Implement missing-braces warning and add a test case. llvm-svn: 97893
2010-03-07 12:17:15 +08:00
#include "clang/Lex/Preprocessor.h"
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
#include "clang/AST/ASTContext.h"
Struggle mightily against header inclusion in Sema.h. llvm-svn: 111904
2010-08-24 15:21:54 +08:00
#include "clang/AST/DeclObjC.h"
Create CXXConstructExprs when constructing via copy initialization. llvm-svn: 72474
2009-05-28 00:10:08 +08:00
#include "clang/AST/ExprCXX.h"
handle @encode interactions with array initializers. llvm-svn: 65401
2009-02-25 07:10:27 +08:00
#include "clang/AST/ExprObjC.h"
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
#include "clang/AST/TypeLoc.h"
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
#include "llvm/Support/ErrorHandling.h"
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
#include <map>
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
using namespace clang;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
//===----------------------------------------------------------------------===//
// Sema Initialization Checking
//===----------------------------------------------------------------------===//
handle @encode interactions with array initializers. llvm-svn: 65401
2009-02-25 07:10:27 +08:00
static Expr *IsStringInit(Expr *Init, QualType DeclType, ASTContext &Context) {
allow wide strings to initialize arrays compatible with wchar_t. Unfortunately this doesn't work yet because wchar_t is completely broken in C. llvm-svn: 65585
2009-02-27 07:26:43 +08:00
const ArrayType *AT = Context.getAsArrayType(DeclType);
if (!AT) return 0;
Revert r72575, which isn't really right, and fix up other code to handle the construct in question correctly. llvm-svn: 72581
2009-05-30 02:22:49 +08:00
if (!isa<ConstantArrayType>(AT) && !isa<IncompleteArrayType>(AT))
return 0;
allow wide strings to initialize arrays compatible with wchar_t. Unfortunately this doesn't work yet because wchar_t is completely broken in C. llvm-svn: 65585
2009-02-27 07:26:43 +08:00
// See if this is a string literal or @encode.
Init = Init->IgnoreParens();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
allow wide strings to initialize arrays compatible with wchar_t. Unfortunately this doesn't work yet because wchar_t is completely broken in C. llvm-svn: 65585
2009-02-27 07:26:43 +08:00
// Handle @encode, which is a narrow string.
if (isa<ObjCEncodeExpr>(Init) && AT->getElementType()->isCharType())
return Init;
// Otherwise we can only handle string literals.
StringLiteral *SL = dyn_cast<StringLiteral>(Init);
fix a bozobug. llvm-svn: 65589
2009-02-27 07:42:47 +08:00
if (SL == 0) return 0;
Fix for PR4285: allow intializing a const wchar_t array with a wide string. llvm-svn: 72663
2009-05-31 18:54:53 +08:00
QualType ElemTy = Context.getCanonicalType(AT->getElementType());
allow wide strings to initialize arrays compatible with wchar_t. Unfortunately this doesn't work yet because wchar_t is completely broken in C. llvm-svn: 65585
2009-02-27 07:26:43 +08:00
// char array can be initialized with a narrow string.
// Only allow char x[] = "foo"; not char x[] = L"foo";
if (!SL->isWide())
Fix for PR4285: allow intializing a const wchar_t array with a wide string. llvm-svn: 72663
2009-05-31 18:54:53 +08:00
return ElemTy->isCharType() ? Init : 0;
allow wide strings to initialize arrays compatible with wchar_t. Unfortunately this doesn't work yet because wchar_t is completely broken in C. llvm-svn: 65585
2009-02-27 07:26:43 +08:00
Fix for PR4285: allow intializing a const wchar_t array with a wide string. llvm-svn: 72663
2009-05-31 18:54:53 +08:00
// wchar_t array can be initialized with a wide string: C99 6.7.8p15 (with
// correction from DR343): "An array with element type compatible with a
// qualified or unqualified version of wchar_t may be initialized by a wide
// string literal, optionally enclosed in braces."
if (Context.typesAreCompatible(Context.getWCharType(),
ElemTy.getUnqualifiedType()))
allow wide strings to initialize arrays compatible with wchar_t. Unfortunately this doesn't work yet because wchar_t is completely broken in C. llvm-svn: 65585
2009-02-27 07:26:43 +08:00
return Init;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
return 0;
}
handle @encode interactions with array initializers. llvm-svn: 65401
2009-02-25 07:10:27 +08:00
static void CheckStringInit(Expr *Str, QualType &DeclT, Sema &S) {
// Get the length of the string as parsed.
uint64_t StrLength =
cast<ConstantArrayType>(Str->getType())->getSize().getZExtValue();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
handle @encode interactions with array initializers. llvm-svn: 65401
2009-02-25 07:10:27 +08:00
const ArrayType *AT = S.Context.getAsArrayType(DeclT);
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
if (const IncompleteArrayType *IAT = dyn_cast<IncompleteArrayType>(AT)) {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
// C99 6.7.8p14. We have an array of character type with unknown size
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
// being initialized to a string literal.
llvm::APSInt ConstVal(32);
rename CheckStringLiteralInit to CheckStringInit and pass in the string size as an argument. llvm-svn: 65400
2009-02-25 07:01:39 +08:00
ConstVal = StrLength;
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
// Return a new array type (C99 6.7.8p22).
Remove the ConstantArrayType subtypes. This information is preserved in the TypeLoc records for declarations; it should not be necessary to represent it directly in the type system. Please complain if you were using these classes and feel you can't replicate previous functionality using the TypeLoc API. llvm-svn: 84222
2009-10-16 08:14:28 +08:00
DeclT = S.Context.getConstantArrayType(IAT->getElementType(),
ConstVal,
ArrayType::Normal, 0);
rename CheckStringLiteralInit to CheckStringInit and pass in the string size as an argument. llvm-svn: 65400
2009-02-25 07:01:39 +08:00
return;
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Revert r72575, which isn't really right, and fix up other code to handle the construct in question correctly. llvm-svn: 72581
2009-05-30 02:22:49 +08:00
const ConstantArrayType *CAT = cast<ConstantArrayType>(AT);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Revert r72575, which isn't really right, and fix up other code to handle the construct in question correctly. llvm-svn: 72581
2009-05-30 02:22:49 +08:00
// C99 6.7.8p14. We have an array of character type with known size. However,
// the size may be smaller or larger than the string we are initializing.
// FIXME: Avoid truncation for 64-bit length strings.
if (StrLength-1 > CAT->getSize().getZExtValue())
S.Diag(Str->getSourceRange().getBegin(),
diag::warn_initializer_string_for_char_array_too_long)
<< Str->getSourceRange();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Revert r72575, which isn't really right, and fix up other code to handle the construct in question correctly. llvm-svn: 72581
2009-05-30 02:22:49 +08:00
// Set the type to the actual size that we are initializing. If we have
// something like:
// char x[1] = "foo";
// then this will set the string literal's type to char[1].
Str->setType(DeclT);
move some initialization checking code from SemaDecl.cpp to SemaInit.cpp, no functionality change. llvm-svn: 65394
2009-02-25 06:27:37 +08:00
}
//===----------------------------------------------------------------------===//
// Semantic checking for initializer lists.
//===----------------------------------------------------------------------===//
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
/// @brief Semantic checking for initializer lists.
///
/// The InitListChecker class contains a set of routines that each
/// handle the initialization of a certain kind of entity, e.g.,
/// arrays, vectors, struct/union types, scalars, etc. The
/// InitListChecker itself performs a recursive walk of the subobject
/// structure of the type to be initialized, while stepping through
/// the initializer list one element at a time. The IList and Index
/// parameters to each of the Check* routines contain the active
/// (syntactic) initializer list and the index into that initializer
/// list that represents the current initializer. Each routine is
/// responsible for moving that Index forward as it consumes elements.
///
/// Each Check* routine also has a StructuredList/StructuredIndex
/// arguments, which contains the current the "structured" (semantic)
/// initializer list and the index into that initializer list where we
/// are copying initializers as we map them over to the semantic
/// list. Once we have completed our recursive walk of the subobject
/// structure, we will have constructed a full semantic initializer
/// list.
///
/// C99 designators cause changes in the initializer list traversal,
/// because they make the initialization "jump" into a specific
/// subobject and then continue the initialization from that
/// point. CheckDesignatedInitializer() recursively steps into the
/// designated subobject and manages backing out the recursion to
/// initialize the subobjects after the one designated.
move InitListChecker to be private to SemaInit.cpp llvm-svn: 65398
2009-02-25 06:48:58 +08:00
namespace {
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
class InitListChecker {
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
Sema &SemaRef;
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
bool hadError;
std::map<InitListExpr *, InitListExpr *> SyntacticToSemantic;
InitListExpr *FullyStructuredList;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckImplicitInitList(const InitializedEntity &Entity,
More init work, adding more entity parameters. llvm-svn: 94332
2010-01-24 04:47:59 +08:00
InitListExpr *ParentIList, QualType T,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
unsigned &Index, InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject = false);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckExplicitInitList(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType &T,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
unsigned &Index, InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject = false);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckListElementTypes(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType &DeclType,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
bool SubobjectIsDesignatorContext,
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
unsigned &Index,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject = false);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckSubElementType(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType ElemType,
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
unsigned &Index,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
InitListExpr *StructuredList,
unsigned &StructuredIndex);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckScalarType(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType DeclType,
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
unsigned &Index,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
InitListExpr *StructuredList,
unsigned &StructuredIndex);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckReferenceType(const InitializedEntity &Entity,
InitListExpr *IList, QualType DeclType,
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
unsigned &Index,
InitListExpr *StructuredList,
unsigned &StructuredIndex);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckVectorType(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType DeclType, unsigned &Index,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
InitListExpr *StructuredList,
unsigned &StructuredIndex);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckStructUnionTypes(const InitializedEntity &Entity,
Use the new init code for member subobjects. llvm-svn: 94329
2010-01-24 04:20:40 +08:00
InitListExpr *IList, QualType DeclType,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
RecordDecl::field_iterator Field,
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
bool SubobjectIsDesignatorContext, unsigned &Index,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject = false);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void CheckArrayType(const InitializedEntity &Entity,
Switch some array initialization over to the new init code. llvm-svn: 94327
2010-01-24 04:13:41 +08:00
InitListExpr *IList, QualType &DeclType,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
llvm::APSInt elementIndex,
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
bool SubobjectIsDesignatorContext, unsigned &Index,
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
InitListExpr *StructuredList,
unsigned &StructuredIndex);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
bool CheckDesignatedInitializer(const InitializedEntity &Entity,
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
InitListExpr *IList, DesignatedInitExpr *DIE,
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
unsigned DesigIdx,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
QualType &CurrentObjectType,
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
RecordDecl::field_iterator *NextField,
llvm::APSInt *NextElementIndex,
unsigned &Index,
InitListExpr *StructuredList,
unsigned &StructuredIndex,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
bool FinishSubobjectInit,
bool TopLevelObject);
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
InitListExpr *getStructuredSubobjectInit(InitListExpr *IList, unsigned Index,
QualType CurrentObjectType,
InitListExpr *StructuredList,
unsigned StructuredIndex,
SourceRange InitRange);
Better documentation for our initialization checker llvm-svn: 63261
2009-01-29 09:05:33 +08:00
void UpdateStructuredListElement(InitListExpr *StructuredList,
unsigned &StructuredIndex,
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
Expr *expr);
int numArrayElements(QualType DeclType);
int numStructUnionElements(QualType DeclType);
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
When filling in value initializations within an initializer list, be sure to fill in the initialized member of a union when a member was explicitly designated. Fixes PR5843. llvm-svn: 91858
2009-12-22 08:05:34 +08:00
void FillInValueInitForField(unsigned Init, FieldDecl *Field,
const InitializedEntity &ParentEntity,
InitListExpr *ILE, bool &RequiresSecondPass);
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
void FillInValueInitializations(const InitializedEntity &Entity,
InitListExpr *ILE, bool &RequiresSecondPass);
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
public:
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
InitListChecker(Sema &S, const InitializedEntity &Entity,
InitListExpr *IL, QualType &T);
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
bool HadError() { return hadError; }
// @brief Retrieves the fully-structured initializer list used for
// semantic analysis and code generation.
InitListExpr *getFullyStructuredList() const { return FullyStructuredList; }
};
move InitListChecker to be private to SemaInit.cpp llvm-svn: 65398
2009-02-25 06:48:58 +08:00
} // end anonymous namespace
fix an absolutely inscrutible gcc 4.0 error: llvm[0]: Compiling SemaInit.cpp for Debug build SemaInit.cpp:171: error: ‘InitListChecker’ has not been declared SemaInit.cpp:171: error: ISO C++ forbids declaration of ‘InitListChecker’ with no type SemaInit.cpp: In function ‘int InitListChecker(clang::Sema*, clang::InitListExpr*, clang::QualType&)’: SemaInit.cpp:172: error: ‘hadError’ was not declared in this scope SemaInit.cpp:173: error: ‘SemaRef’ was not declared in this scope SemaInit.cpp:177: error: ‘FullyStructuredList’ was not declared in this scope llvm-svn: 63270
2009-01-29 13:10:57 +08:00
When filling in value initializations within an initializer list, be sure to fill in the initialized member of a union when a member was explicitly designated. Fixes PR5843. llvm-svn: 91858
2009-12-22 08:05:34 +08:00
void InitListChecker::FillInValueInitForField(unsigned Init, FieldDecl *Field,
const InitializedEntity &ParentEntity,
InitListExpr *ILE,
bool &RequiresSecondPass) {
SourceLocation Loc = ILE->getSourceRange().getBegin();
unsigned NumInits = ILE->getNumInits();
InitializedEntity MemberEntity
= InitializedEntity::InitializeMember(Field, &ParentEntity);
if (Init >= NumInits || !ILE->getInit(Init)) {
// FIXME: We probably don't need to handle references
// specially here, since value-initialization of references is
// handled in InitializationSequence.
if (Field->getType()->isReferenceType()) {
// C++ [dcl.init.aggr]p9:
// If an incomplete or empty initializer-list leaves a
// member of reference type uninitialized, the program is
// ill-formed.
SemaRef.Diag(Loc, diag::err_init_reference_member_uninitialized)
<< Field->getType()
<< ILE->getSyntacticForm()->getSourceRange();
SemaRef.Diag(Field->getLocation(),
diag::note_uninit_reference_member);
hadError = true;
return;
}
InitializationKind Kind = InitializationKind::CreateValue(Loc, Loc, Loc,
true);
InitializationSequence InitSeq(SemaRef, MemberEntity, Kind, 0, 0);
if (!InitSeq) {
InitSeq.Diagnose(SemaRef, MemberEntity, Kind, 0, 0);
hadError = true;
return;
}
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult MemberInit
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
= InitSeq.Perform(SemaRef, MemberEntity, Kind, MultiExprArg());
When filling in value initializations within an initializer list, be sure to fill in the initialized member of a union when a member was explicitly designated. Fixes PR5843. llvm-svn: 91858
2009-12-22 08:05:34 +08:00
if (MemberInit.isInvalid()) {
hadError = true;
return;
}
if (hadError) {
// Do nothing
} else if (Init < NumInits) {
ILE->setInit(Init, MemberInit.takeAs<Expr>());
} else if (InitSeq.getKind()
== InitializationSequence::ConstructorInitialization) {
// Value-initialization requires a constructor call, so
// extend the initializer list to include the constructor
// call and make a note that we'll need to take another pass
// through the initializer list.
Use ASTVector instead of std::vector for the Exprs in InitListExpr. Performance measurements of '-fsyntax-only' on combine.c (403.gcc) shows no real performance change, but now the vector isn't leaked. llvm-svn: 101195
2010-04-14 07:39:13 +08:00
ILE->updateInit(SemaRef.Context, Init, MemberInit.takeAs<Expr>());
When filling in value initializations within an initializer list, be sure to fill in the initialized member of a union when a member was explicitly designated. Fixes PR5843. llvm-svn: 91858
2009-12-22 08:05:34 +08:00
RequiresSecondPass = true;
}
} else if (InitListExpr *InnerILE
= dyn_cast<InitListExpr>(ILE->getInit(Init)))
FillInValueInitializations(MemberEntity, InnerILE,
RequiresSecondPass);
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
/// Recursively replaces NULL values within the given initializer list
/// with expressions that perform value-initialization of the
/// appropriate type.
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
void
InitListChecker::FillInValueInitializations(const InitializedEntity &Entity,
InitListExpr *ILE,
bool &RequiresSecondPass) {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
assert((ILE->getType() != SemaRef.Context.VoidTy) &&
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
"Should not have void type");
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
SourceLocation Loc = ILE->getSourceRange().getBegin();
if (ILE->getSyntacticForm())
Loc = ILE->getSyntacticForm()->getSourceRange().getBegin();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
if (const RecordType *RType = ILE->getType()->getAs<RecordType>()) {
When filling in value initializations within an initializer list, be sure to fill in the initialized member of a union when a member was explicitly designated. Fixes PR5843. llvm-svn: 91858
2009-12-22 08:05:34 +08:00
if (RType->getDecl()->isUnion() &&
ILE->getInitializedFieldInUnion())
FillInValueInitForField(0, ILE->getInitializedFieldInUnion(),
Entity, ILE, RequiresSecondPass);
else {
unsigned Init = 0;
for (RecordDecl::field_iterator
Field = RType->getDecl()->field_begin(),
FieldEnd = RType->getDecl()->field_end();
Field != FieldEnd; ++Field) {
if (Field->isUnnamedBitfield())
continue;
if (hadError)
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
return;
When filling in value initializations within an initializer list, be sure to fill in the initialized member of a union when a member was explicitly designated. Fixes PR5843. llvm-svn: 91858
2009-12-22 08:05:34 +08:00
FillInValueInitForField(Init, *Field, Entity, ILE, RequiresSecondPass);
if (hadError)
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
return;
When filling in value initializations within an initializer list, be sure to fill in the initialized member of a union when a member was explicitly designated. Fixes PR5843. llvm-svn: 91858
2009-12-22 08:05:34 +08:00
++Init;
// Only look at the first initialization of a union.
if (RType->getDecl()->isUnion())
break;
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
}
return;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
QualType ElementType;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
InitializedEntity ElementEntity = Entity;
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
unsigned NumInits = ILE->getNumInits();
unsigned NumElements = NumInits;
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
if (const ArrayType *AType = SemaRef.Context.getAsArrayType(ILE->getType())) {
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
ElementType = AType->getElementType();
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
if (const ConstantArrayType *CAType = dyn_cast<ConstantArrayType>(AType))
NumElements = CAType->getSize().getZExtValue();
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
ElementEntity = InitializedEntity::InitializeElement(SemaRef.Context,
0, Entity);
Change all the Type::getAsFoo() methods to specializations of Type::getAs(). Several of the existing methods were identical to their respective specializations, and so have been removed entirely. Several more 'leaf' optimizations were introduced. The getAsFoo() methods which imposed extra conditions, like getAsObjCInterfacePointerType(), have been left in place. llvm-svn: 82501
2009-09-22 07:43:11 +08:00
} else if (const VectorType *VType = ILE->getType()->getAs<VectorType>()) {
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
ElementType = VType->getElementType();
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
NumElements = VType->getNumElements();
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
ElementEntity = InitializedEntity::InitializeElement(SemaRef.Context,
0, Entity);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
} else
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
ElementType = ILE->getType();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
for (unsigned Init = 0; Init != NumElements; ++Init) {
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
if (hadError)
return;
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
if (ElementEntity.getKind() == InitializedEntity::EK_ArrayElement ||
ElementEntity.getKind() == InitializedEntity::EK_VectorElement)
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
ElementEntity.setElementIndex(Init);
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
if (Init >= NumInits || !ILE->getInit(Init)) {
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
InitializationKind Kind = InitializationKind::CreateValue(Loc, Loc, Loc,
true);
InitializationSequence InitSeq(SemaRef, ElementEntity, Kind, 0, 0);
if (!InitSeq) {
InitSeq.Diagnose(SemaRef, ElementEntity, Kind, 0, 0);
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
hadError = true;
return;
}
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult ElementInit
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
= InitSeq.Perform(SemaRef, ElementEntity, Kind, MultiExprArg());
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
if (ElementInit.isInvalid()) {
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
hadError = true;
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
return;
}
if (hadError) {
// Do nothing
} else if (Init < NumInits) {
ILE->setInit(Init, ElementInit.takeAs<Expr>());
} else if (InitSeq.getKind()
== InitializationSequence::ConstructorInitialization) {
// Value-initialization requires a constructor call, so
// extend the initializer list to include the constructor
// call and make a note that we'll need to take another pass
// through the initializer list.
Use ASTVector instead of std::vector for the Exprs in InitListExpr. Performance measurements of '-fsyntax-only' on combine.c (403.gcc) shows no real performance change, but now the vector isn't leaked. llvm-svn: 101195
2010-04-14 07:39:13 +08:00
ILE->updateInit(SemaRef.Context, Init, ElementInit.takeAs<Expr>());
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
RequiresSecondPass = true;
}
Canonicalize else. llvm-svn: 78102
2009-08-05 05:02:39 +08:00
} else if (InitListExpr *InnerILE
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
= dyn_cast<InitListExpr>(ILE->getInit(Init)))
FillInValueInitializations(ElementEntity, InnerILE, RequiresSecondPass);
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
}
}
fix an absolutely inscrutible gcc 4.0 error: llvm[0]: Compiling SemaInit.cpp for Debug build SemaInit.cpp:171: error: ‘InitListChecker’ has not been declared SemaInit.cpp:171: error: ISO C++ forbids declaration of ‘InitListChecker’ with no type SemaInit.cpp: In function ‘int InitListChecker(clang::Sema*, clang::InitListExpr*, clang::QualType&)’: SemaInit.cpp:172: error: ‘hadError’ was not declared in this scope SemaInit.cpp:173: error: ‘SemaRef’ was not declared in this scope SemaInit.cpp:177: error: ‘FullyStructuredList’ was not declared in this scope llvm-svn: 63270
2009-01-29 13:10:57 +08:00
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
InitListChecker::InitListChecker(Sema &S, const InitializedEntity &Entity,
InitListExpr *IL, QualType &T)
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
: SemaRef(S) {
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
hadError = false;
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
unsigned newIndex = 0;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned newStructuredIndex = 0;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
FullyStructuredList
Fix PR3509 by providing correct starting locations for initializer lists llvm-svn: 65777
2009-03-02 01:12:46 +08:00
= getStructuredSubobjectInit(IL, newIndex, T, 0, 0, IL->getSourceRange());
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
CheckExplicitInitList(Entity, IL, T, newIndex,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
FullyStructuredList, newStructuredIndex,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
/*TopLevelObject=*/true);
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
if (!hadError) {
bool RequiresSecondPass = false;
FillInValueInitializations(Entity, FullyStructuredList, RequiresSecondPass);
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
if (RequiresSecondPass && !hadError)
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
FillInValueInitializations(Entity, FullyStructuredList,
RequiresSecondPass);
}
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
int InitListChecker::numArrayElements(QualType DeclType) {
Move the error checking for variable-sized objects so we don't double-report errors; fixes PR2362. llvm-svn: 51555
2008-05-25 21:22:35 +08:00
// FIXME: use a proper constant
int maxElements = 0x7FFFFFFF;
Finally fix PR2189. This makes a fairly invasive but important change to move getAsArrayType into ASTContext instead of being a method on type. This is required because getAsArrayType(const AT), where AT is a typedef for "int[10]" needs to return ArrayType(const int, 10). Fixing this greatly simplifies getArrayDecayedType, which is a good sign. llvm-svn: 54317
2008-08-04 15:31:14 +08:00
if (const ConstantArrayType *CAT =
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Context.getAsConstantArrayType(DeclType)) {
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
maxElements = static_cast<int>(CAT->getSize().getZExtValue());
}
return maxElements;
}
int InitListChecker::numStructUnionElements(QualType DeclType) {
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
RecordDecl *structDecl = DeclType->getAs<RecordType>()->getDecl();
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
int InitializableMembers = 0;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
for (RecordDecl::field_iterator
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
Field = structDecl->field_begin(),
FieldEnd = structDecl->field_end();
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
Field != FieldEnd; ++Field) {
if ((*Field)->getIdentifier() || !(*Field)->isBitField())
++InitializableMembers;
}
-Changes to TagDecl: Added TagKind enum. Added getTagKind() method. Added convenience methods: isEnum(), isStruct(), isUnion(), isClass(). -RecordDecl/CXXRecordDecl::Create() accept a TagKind enum instead of a DeclKind one. llvm-svn: 52160
2008-06-10 07:19:58 +08:00
if (structDecl->isUnion())
Count the number of initializable members correctly in structs/unions with unnamed members. llvm-svn: 51557
2008-05-25 22:03:31 +08:00
return std::min(InitializableMembers, 1);
return InitializableMembers - structDecl->hasFlexibleArrayMember();
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckImplicitInitList(const InitializedEntity &Entity,
More init work, adding more entity parameters. llvm-svn: 94332
2010-01-24 04:47:59 +08:00
InitListExpr *ParentIList,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
QualType T, unsigned &Index,
InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject) {
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
int maxElements = 0;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
if (T->isArrayType())
maxElements = numArrayElements(T);
Introduce Type::isStructureOrClassType(), which does the obvious thing. Audit all uses of Type::isStructure(), changing those calls to isStructureOrClassType() as needed (which is alsmost everywhere). Fixes the remaining failure in Boost.Utility/Swap. llvm-svn: 102386
2010-04-27 05:31:17 +08:00
else if (T->isRecordType())
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
maxElements = numStructUnionElements(T);
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
else if (T->isVectorType())
Change all the Type::getAsFoo() methods to specializations of Type::getAs(). Several of the existing methods were identical to their respective specializations, and so have been removed entirely. Several more 'leaf' optimizations were introduced. The getAsFoo() methods which imposed extra conditions, like getAsObjCInterfacePointerType(), have been left in place. llvm-svn: 82501
2009-09-22 07:43:11 +08:00
maxElements = T->getAs<VectorType>()->getNumElements();
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
else
assert(0 && "CheckImplicitInitList(): Illegal type");
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
Diagnose implicit init list for empty aggregate, like struct {}. Fixes PR2151 (by not creating the empty implicit init list). llvm-svn: 51556
2008-05-25 21:49:22 +08:00
if (maxElements == 0) {
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(ParentIList->getInit(Index)->getLocStart(),
Diagnose implicit init list for empty aggregate, like struct {}. Fixes PR2151 (by not creating the empty implicit init list). llvm-svn: 51556
2008-05-25 21:49:22 +08:00
diag::err_implicit_empty_initializer);
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++Index;
Diagnose implicit init list for empty aggregate, like struct {}. Fixes PR2151 (by not creating the empty implicit init list). llvm-svn: 51556
2008-05-25 21:49:22 +08:00
hadError = true;
return;
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Build a structured initializer list corresponding to this subobject.
InitListExpr *StructuredSubobjectInitList
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
= getStructuredSubobjectInit(ParentIList, Index, T, StructuredList,
StructuredIndex,
Fix PR3509 by providing correct starting locations for initializer lists llvm-svn: 65777
2009-03-02 01:12:46 +08:00
SourceRange(ParentIList->getInit(Index)->getSourceRange().getBegin(),
ParentIList->getSourceRange().getEnd()));
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned StructuredSubobjectInitIndex = 0;
// Check the element types and build the structural subobject.
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
unsigned StartIndex = Index;
More init work, adding more entity parameters. llvm-svn: 94332
2010-01-24 04:47:59 +08:00
CheckListElementTypes(Entity, ParentIList, T,
/*SubobjectIsDesignatorContext=*/false, Index,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
StructuredSubobjectInitList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
StructuredSubobjectInitIndex,
TopLevelObject);
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
unsigned EndIndex = (Index == StartIndex? StartIndex : Index - 1);
Allow flexible array initializers that are not surrounded by braces. We now build the appropriate fully-structured initializer list for such things. Per PR3618, verified that we're getting the right code generation. llvm-svn: 67353
2009-03-20 08:32:56 +08:00
StructuredSubobjectInitList->setType(T);
Fix PR3509 by providing correct starting locations for initializer lists llvm-svn: 65777
2009-03-02 01:12:46 +08:00
// Update the structured sub-object initializer so that it's ending
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
// range corresponds with the end of the last initializer it used.
if (EndIndex < ParentIList->getNumInits()) {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SourceLocation EndLoc
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
= ParentIList->getInit(EndIndex)->getSourceRange().getEnd();
StructuredSubobjectInitList->setRBraceLoc(EndLoc);
}
Implement missing-braces warning and add a test case. llvm-svn: 97893
2010-03-07 12:17:15 +08:00
// Warn about missing braces.
if (T->isArrayType() || T->isRecordType()) {
Fix indentation, use string directly instead of StringRef. llvm-svn: 97896
2010-03-07 12:40:06 +08:00
SemaRef.Diag(StructuredSubobjectInitList->getLocStart(),
diag::warn_missing_braces)
Implement missing-braces warning and add a test case. llvm-svn: 97893
2010-03-07 12:17:15 +08:00
<< StructuredSubobjectInitList->getSourceRange()
Reinstate my CodeModificationHint -> FixItHint renaming patch, without the C-only "optimization". llvm-svn: 100022
2010-04-01 01:46:05 +08:00
<< FixItHint::CreateInsertion(StructuredSubobjectInitList->getLocStart(),
"{")
<< FixItHint::CreateInsertion(SemaRef.PP.getLocForEndOfToken(
Fix 80 col violation. llvm-svn: 97898
2010-03-07 12:47:12 +08:00
StructuredSubobjectInitList->getLocEnd()),
Reinstate my CodeModificationHint -> FixItHint renaming patch, without the C-only "optimization". llvm-svn: 100022
2010-04-01 01:46:05 +08:00
"}");
Implement missing-braces warning and add a test case. llvm-svn: 97893
2010-03-07 12:17:15 +08:00
}
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckExplicitInitList(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType &T,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned &Index,
InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject) {
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
assert(IList->isExplicit() && "Illegal Implicit InitListExpr");
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
SyntacticToSemantic[IList] = StructuredList;
StructuredList->setSyntacticForm(IList);
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
CheckListElementTypes(Entity, IList, T, /*SubobjectIsDesignatorContext=*/true,
Index, StructuredList, StructuredIndex, TopLevelObject);
Whenever we're creating an expression that is typically an rvalue (e.g., a call, cast, etc.), immediately adjust the expression's type to strip cv-qualifiers off of all non-class types (in C++) or all types (in C). This effectively extends my previous fix for PR7463, which was restricted to calls, to other kinds of expressions within similar characteristics. I've audited every use of getNonReferenceType() in the code base, switching to the newly-renamed getNonLValueExprType() where necessary. Big thanks to Eli for pointing out just how incomplete my original fix for PR7463 actually was. We've been handling cv-qualifiers on rvalues wrong for a very, very long time. Fixes PR7463. llvm-svn: 108253
2010-07-14 02:40:04 +08:00
QualType ExprTy = T.getNonLValueExprType(SemaRef.Context);
IList->setType(ExprTy);
StructuredList->setType(ExprTy);
Move the error checking for variable-sized objects so we don't double-report errors; fixes PR2362. llvm-svn: 51555
2008-05-25 21:22:35 +08:00
if (hadError)
return;
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
Move the error checking for variable-sized objects so we don't double-report errors; fixes PR2362. llvm-svn: 51555
2008-05-25 21:22:35 +08:00
if (Index < IList->getNumInits()) {
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
// We have leftover initializers
Make sure we don't give the wrong warning, and make sure not to set hadError (suppressing future diagnostics) if we didn't print an error. llvm-svn: 72588
2009-05-30 04:20:05 +08:00
if (StructuredIndex == 1 &&
IsStringInit(StructuredList->getInit(0), T, SemaRef.Context)) {
Downgrade the "excess elements in initializer" errors to warnings *in C*. They're required errors in C++. llvm-svn: 64964
2009-02-19 06:23:55 +08:00
unsigned DK = diag::warn_excess_initializers_in_char_array_initializer;
Make sure we don't give the wrong warning, and make sure not to set hadError (suppressing future diagnostics) if we didn't print an error. llvm-svn: 72588
2009-05-30 04:20:05 +08:00
if (SemaRef.getLangOptions().CPlusPlus) {
Downgrade the "excess elements in initializer" errors to warnings *in C*. They're required errors in C++. llvm-svn: 64964
2009-02-19 06:23:55 +08:00
DK = diag::err_excess_initializers_in_char_array_initializer;
Make sure we don't give the wrong warning, and make sure not to set hadError (suppressing future diagnostics) if we didn't print an error. llvm-svn: 72588
2009-05-30 04:20:05 +08:00
hadError = true;
}
Add errors for some illegal constructs (specifically, "int a = {{3}};" and "int a = {};"). I'll adjust the tests in a bit. llvm-svn: 51265
2008-05-20 04:12:18 +08:00
// Special-case
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getInit(Index)->getLocStart(), DK)
remove one more old-style Diag method. llvm-svn: 59589
2008-11-19 13:27:50 +08:00
<< IList->getInit(Index)->getSourceRange();
Be a bit more defensive in SemaInit. llvm-svn: 51301
2008-05-20 13:25:56 +08:00
} else if (!T->isIncompleteType()) {
Upgrade the "excess elements in array initializer" warning to an error, since both C99 and C++ consider it an error. For reference, GCC makes this a warning while G++ makes it an error. llvm-svn: 63435
2009-01-31 06:26:29 +08:00
// Don't complain for incomplete types, since we'll get an error
// elsewhere
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
QualType CurrentObjectType = StructuredList->getType();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
int initKind =
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
CurrentObjectType->isArrayType()? 0 :
CurrentObjectType->isVectorType()? 1 :
CurrentObjectType->isScalarType()? 2 :
CurrentObjectType->isUnionType()? 3 :
4;
Downgrade the "excess elements in initializer" errors to warnings *in C*. They're required errors in C++. llvm-svn: 64964
2009-02-19 06:23:55 +08:00
unsigned DK = diag::warn_excess_initializers;
Make sure we don't give the wrong warning, and make sure not to set hadError (suppressing future diagnostics) if we didn't print an error. llvm-svn: 72588
2009-05-30 04:20:05 +08:00
if (SemaRef.getLangOptions().CPlusPlus) {
DK = diag::err_excess_initializers;
Extra vector element initializers in OpenCL is an error, not a warning. llvm-svn: 74951
2009-07-08 05:53:06 +08:00
hadError = true;
}
if (SemaRef.getLangOptions().OpenCL && initKind == 1) {
DK = diag::err_excess_initializers;
Make sure we don't give the wrong warning, and make sure not to set hadError (suppressing future diagnostics) if we didn't print an error. llvm-svn: 72588
2009-05-30 04:20:05 +08:00
hadError = true;
}
Downgrade the "excess elements in initializer" errors to warnings *in C*. They're required errors in C++. llvm-svn: 64964
2009-02-19 06:23:55 +08:00
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getInit(Index)->getLocStart(), DK)
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
<< initKind << IList->getInit(Index)->getSourceRange();
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
}
}
Adjust warning so that it doesn't fire when there is an error. llvm-svn: 51269
2008-05-20 04:20:43 +08:00
PR3009: Get rid of bogus warning for scalar compound literals. This patch isn't quite ideal in that it eliminates the warning for constructs like "int a = {1};", where the braces are in fact redundant. However, that would have required a bunch of refactoring, and it's much less likely to cause confusion compared to redundant nested braces. llvm-svn: 71939
2009-05-16 19:45:48 +08:00
if (T->isScalarType() && !TopLevelObject)
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getLocStart(), diag::warn_braces_around_scalar_init)
Add some more code modification hints llvm-svn: 68261
2009-04-02 07:51:29 +08:00
<< IList->getSourceRange()
Reinstate my CodeModificationHint -> FixItHint renaming patch, without the C-only "optimization". llvm-svn: 100022
2010-04-01 01:46:05 +08:00
<< FixItHint::CreateRemoval(IList->getLocStart())
<< FixItHint::CreateRemoval(IList->getLocEnd());
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckListElementTypes(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
QualType &DeclType,
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
bool SubobjectIsDesignatorContext,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned &Index,
InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject) {
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
if (DeclType->isScalarType()) {
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
CheckScalarType(Entity, IList, DeclType, Index,
StructuredList, StructuredIndex);
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
} else if (DeclType->isVectorType()) {
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
CheckVectorType(Entity, IList, DeclType, Index,
StructuredList, StructuredIndex);
Switch Type::isAggregateType to use the C++ definition of "aggregate type" rather than the C definition. We do this because both C99 and Clang always use "aggregate type" as "aggregate or union type", and the C++ definition includes union types. llvm-svn: 63395
2009-01-31 01:31:00 +08:00
} else if (DeclType->isAggregateType()) {
if (DeclType->isRecordType()) {
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
RecordDecl *RD = DeclType->getAs<RecordType>()->getDecl();
Use the new init code for member subobjects. llvm-svn: 94329
2010-01-24 04:20:40 +08:00
CheckStructUnionTypes(Entity, IList, DeclType, RD->field_begin(),
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
SubobjectIsDesignatorContext, Index,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
StructuredList, StructuredIndex,
TopLevelObject);
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
} else if (DeclType->isArrayType()) {
Properly manage the bit-widths of APInts/APSInts in array initialization. Fixes PR clang/3377 llvm-svn: 62851
2009-01-24 00:54:12 +08:00
llvm::APSInt Zero(
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Context.getTypeSize(SemaRef.Context.getSizeType()),
Properly manage the bit-widths of APInts/APSInts in array initialization. Fixes PR clang/3377 llvm-svn: 62851
2009-01-24 00:54:12 +08:00
false);
Switch some array initialization over to the new init code. llvm-svn: 94327
2010-01-24 04:13:41 +08:00
CheckArrayType(Entity, IList, DeclType, Zero,
SubobjectIsDesignatorContext, Index,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
StructuredList, StructuredIndex);
Canonicalize else. llvm-svn: 78102
2009-08-05 05:02:39 +08:00
} else
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
assert(0 && "Aggregate that isn't a structure or array?!");
InitListChecker::CheckListElementTypes(): Check for function types and issue an appropriate diagnostic. llvm-svn: 54614
2008-08-11 00:05:48 +08:00
} else if (DeclType->isVoidType() || DeclType->isFunctionType()) {
// This type is invalid, issue a diagnostic.
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++Index;
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getLocStart(), diag::err_illegal_initializer_type)
Change a whole lot of diagnostics to take QualType's directly instead of converting them to strings first. This also fixes a bunch of minor inconsistencies in the diagnostics emitted by clang and adds a bunch of FIXME's to DiagnosticKinds.def. llvm-svn: 59948
2008-11-24 14:25:27 +08:00
<< DeclType;
Be a bit more defensive in SemaInit. llvm-svn: 51301
2008-05-20 13:25:56 +08:00
hadError = true;
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
} else if (DeclType->isRecordType()) {
// C++ [dcl.init]p14:
// [...] If the class is an aggregate (8.5.1), and the initializer
// is a brace-enclosed list, see 8.5.1.
//
// Note: 8.5.1 is handled below; here, we diagnose the case where
// we have an initializer list and a destination type that is not
// an aggregate.
// FIXME: In C++0x, this is yet another form of initialization.
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getLocStart(), diag::err_init_non_aggr_init_list)
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
<< DeclType << IList->getSourceRange();
hadError = true;
} else if (DeclType->isReferenceType()) {
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
CheckReferenceType(Entity, IList, DeclType, Index,
StructuredList, StructuredIndex);
Substantially alter the design of the Objective C type AST by introducing ObjCObjectType, which is basically just a pair of one of {primitive-id, primitive-Class, user-defined @class} with a list of protocols. An ObjCObjectPointerType is therefore just a pointer which always points to one of these types (possibly sugared). ObjCInterfaceType is now just a kind of ObjCObjectType which happens to not carry any protocols. Alter a rather large number of use sites to use ObjCObjectType instead of ObjCInterfaceType. Store an ObjCInterfaceType as a pointer on the decl rather than hashing them in a FoldingSet. Remove some number of methods that are no longer used, at least after this patch. By simplifying ObjCObjectPointerType, we are now able to easily remove and apply pointers to Objective-C types, which is crucial for a certain kind of ObjC++ metaprogramming common in WebKit. llvm-svn: 103870
2010-05-15 19:32:37 +08:00
} else if (DeclType->isObjCObjectType()) {
Complain when we try to initialize an object of Objective-C class type (which is ill-formed) with an initializer list. Also, change the fallback from an assertion to a generic error message, which is far friendlier. Fixes <rdar://problem/7730948>. llvm-svn: 102930
2010-05-04 02:24:37 +08:00
SemaRef.Diag(IList->getLocStart(), diag::err_init_objc_class)
<< DeclType;
hadError = true;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
} else {
Complain when we try to initialize an object of Objective-C class type (which is ill-formed) with an initializer list. Also, change the fallback from an assertion to a generic error message, which is far friendlier. Fixes <rdar://problem/7730948>. llvm-svn: 102930
2010-05-04 02:24:37 +08:00
SemaRef.Diag(IList->getLocStart(), diag::err_illegal_initializer_type)
<< DeclType;
hadError = true;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckSubElementType(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
QualType ElemType,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned &Index,
InitListExpr *StructuredList,
unsigned &StructuredIndex) {
Eliminate infinite looping in a wacky case with designated initializers. Simplifies (somewhat) the actually checking of the initializer expression following the designators llvm-svn: 63257
2009-01-29 08:39:20 +08:00
Expr *expr = IList->getInit(Index);
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
if (InitListExpr *SubInitList = dyn_cast<InitListExpr>(expr)) {
unsigned newIndex = 0;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned newStructuredIndex = 0;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
InitListExpr *newStructuredList
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
= getStructuredSubobjectInit(IList, Index, ElemType,
StructuredList, StructuredIndex,
SubInitList->getSourceRange());
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
CheckExplicitInitList(Entity, SubInitList, ElemType, newIndex,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
newStructuredList, newStructuredIndex);
++StructuredIndex;
++Index;
handle @encode interactions with array initializers. llvm-svn: 65401
2009-02-25 07:10:27 +08:00
} else if (Expr *Str = IsStringInit(expr, ElemType, SemaRef.Context)) {
CheckStringInit(Str, ElemType, SemaRef);
make CheckStringLiteralInit a static function in SemaInit.cpp llvm-svn: 65396
2009-02-25 06:41:04 +08:00
UpdateStructuredListElement(StructuredList, StructuredIndex, Str);
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++Index;
Some more SemaInit cleanup. llvm-svn: 51262
2008-05-20 04:00:43 +08:00
} else if (ElemType->isScalarType()) {
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
CheckScalarType(Entity, IList, ElemType, Index,
StructuredList, StructuredIndex);
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
} else if (ElemType->isReferenceType()) {
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
CheckReferenceType(Entity, IList, ElemType, Index,
StructuredList, StructuredIndex);
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
} else {
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
if (SemaRef.getLangOptions().CPlusPlus) {
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
// C++ [dcl.init.aggr]p12:
// All implicit type conversions (clause 4) are considered when
// initializing the aggregate member with an ini- tializer from
// an initializer-list. If the initializer can initialize a
// member, the member is initialized. [...]
Use new initialization code when dealing with [dcl.init.aggr]p12. This fixes the bug where array elements and member initializers weren't copied correctly. llvm-svn: 94340
2010-01-24 08:19:41 +08:00
// FIXME: Better EqualLoc?
InitializationKind Kind =
InitializationKind::CreateCopy(expr->getLocStart(), SourceLocation());
InitializationSequence Seq(SemaRef, Entity, Kind, &expr, 1);
if (Seq) {
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult Result =
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
Seq.Perform(SemaRef, Entity, Kind, MultiExprArg(&expr, 1));
Use new initialization code when dealing with [dcl.init.aggr]p12. This fixes the bug where array elements and member initializers weren't copied correctly. llvm-svn: 94340
2010-01-24 08:19:41 +08:00
if (Result.isInvalid())
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
hadError = true;
Use new initialization code when dealing with [dcl.init.aggr]p12. This fixes the bug where array elements and member initializers weren't copied correctly. llvm-svn: 94340
2010-01-24 08:19:41 +08:00
UpdateStructuredListElement(StructuredList, StructuredIndex,
Result.takeAs<Expr>());
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
++Index;
return;
}
// Fall through for subaggregate initialization
} else {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
// C99 6.7.8p13:
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
//
// The initializer for a structure or union object that has
// automatic storage duration shall be either an initializer
// list as described below, or a single expression that has
// compatible structure or union type. In the latter case, the
// initial value of the object, including unnamed members, is
// that of the expression.
Allow initializing a vector with a vector in addition to allowing a list of the elements. Issue reported on cfe-dev by Mattias Holm. llvm-svn: 73292
2009-06-13 18:38:46 +08:00
if ((ElemType->isRecordType() || ElemType->isVectorType()) &&
Revert r72575, which isn't really right, and fix up other code to handle the construct in question correctly. llvm-svn: 72581
2009-05-30 02:22:49 +08:00
SemaRef.Context.hasSameUnqualifiedType(expr->getType(), ElemType)) {
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
UpdateStructuredListElement(StructuredList, StructuredIndex, expr);
++Index;
return;
}
// Fall through for subaggregate initialization
}
// C++ [dcl.init.aggr]p12:
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
//
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
// [...] Otherwise, if the member is itself a non-empty
// subaggregate, brace elision is assumed and the initializer is
// considered for the initialization of the first member of
// the subaggregate.
if (ElemType->isAggregateType() || ElemType->isVectorType()) {
More init work, adding more entity parameters. llvm-svn: 94332
2010-01-24 04:47:59 +08:00
CheckImplicitInitList(Entity, IList, ElemType, Index, StructuredList,
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
StructuredIndex);
++StructuredIndex;
} else {
// We cannot initialize this element, so let
// PerformCopyInitialization produce the appropriate diagnostic.
Eliminate yet another old-school PerformCopyInitialization. llvm-svn: 94874
2010-01-30 09:56:32 +08:00
SemaRef.PerformCopyInitialization(Entity, SourceLocation(),
SemaRef.Owned(expr));
IList->setInit(Index, 0);
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
hadError = true;
++Index;
++StructuredIndex;
}
}
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckScalarType(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType DeclType,
Eliminate infinite looping in a wacky case with designated initializers. Simplifies (somewhat) the actually checking of the initializer expression following the designators llvm-svn: 63257
2009-01-29 08:39:20 +08:00
unsigned &Index,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
InitListExpr *StructuredList,
unsigned &StructuredIndex) {
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
if (Index < IList->getNumInits()) {
Eliminate infinite looping in a wacky case with designated initializers. Simplifies (somewhat) the actually checking of the initializer expression following the designators llvm-svn: 63257
2009-01-29 08:39:20 +08:00
Expr *expr = IList->getInit(Index);
PR3344: Downgrade "too many braces around scalar initializer" to a warning. llvm-svn: 111067
2010-08-14 11:14:53 +08:00
if (InitListExpr *SubIList = dyn_cast<InitListExpr>(expr)) {
SemaRef.Diag(SubIList->getLocStart(),
diag::warn_many_braces_around_scalar_init)
<< SubIList->getSourceRange();
CheckScalarType(Entity, SubIList, DeclType, Index, StructuredList,
StructuredIndex);
Add errors for some illegal constructs (specifically, "int a = {{3}};" and "int a = {};"). I'll adjust the tests in a bit. llvm-svn: 51265
2008-05-20 04:12:18 +08:00
return;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
} else if (isa<DesignatedInitExpr>(expr)) {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(expr->getSourceRange().getBegin(),
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
diag::err_designator_for_scalar_init)
<< DeclType << expr->getSourceRange();
hadError = true;
++Index;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++StructuredIndex;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
return;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult Result =
Code simplification; this should have no visible effects. llvm-svn: 94427
2010-01-26 01:04:54 +08:00
SemaRef.PerformCopyInitialization(Entity, expr->getLocStart(),
SemaRef.Owned(expr));
Change CheckSingleInitializer to take/return OwningExprResults instead. No functionality change. llvm-svn: 94316
2010-01-24 02:35:41 +08:00
Silence a GCC warning about a possibly uninitialized variable. It's data flow only flows so far it seems. llvm-svn: 96085
2010-02-13 15:23:01 +08:00
Expr *ResultExpr = 0;
Change CheckSingleInitializer to take/return OwningExprResults instead. No functionality change. llvm-svn: 94316
2010-01-24 02:35:41 +08:00
if (Result.isInvalid())
Add errors for some illegal constructs (specifically, "int a = {{3}};" and "int a = {};"). I'll adjust the tests in a bit. llvm-svn: 51265
2008-05-20 04:12:18 +08:00
hadError = true; // types weren't compatible.
Change CheckSingleInitializer to take/return OwningExprResults instead. No functionality change. llvm-svn: 94316
2010-01-24 02:35:41 +08:00
else {
ResultExpr = Result.takeAs<Expr>();
if (ResultExpr != expr) {
// The type was promoted, update initializer list.
IList->setInit(Index, ResultExpr);
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
if (hadError)
++StructuredIndex;
else
Change CheckSingleInitializer to take/return OwningExprResults instead. No functionality change. llvm-svn: 94316
2010-01-24 02:35:41 +08:00
UpdateStructuredListElement(StructuredList, StructuredIndex, ResultExpr);
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
++Index;
Add errors for some illegal constructs (specifically, "int a = {{3}};" and "int a = {};"). I'll adjust the tests in a bit. llvm-svn: 51265
2008-05-20 04:12:18 +08:00
} else {
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getLocStart(), diag::err_empty_scalar_initializer)
remove one more old-style Diag method. llvm-svn: 59589
2008-11-19 13:27:50 +08:00
<< IList->getSourceRange();
Add errors for some illegal constructs (specifically, "int a = {{3}};" and "int a = {};"). I'll adjust the tests in a bit. llvm-svn: 51265
2008-05-20 04:12:18 +08:00
hadError = true;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++Index;
++StructuredIndex;
Add errors for some illegal constructs (specifically, "int a = {{3}};" and "int a = {};"). I'll adjust the tests in a bit. llvm-svn: 51265
2008-05-20 04:12:18 +08:00
return;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckReferenceType(const InitializedEntity &Entity,
InitListExpr *IList, QualType DeclType,
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
unsigned &Index,
InitListExpr *StructuredList,
unsigned &StructuredIndex) {
if (Index < IList->getNumInits()) {
Expr *expr = IList->getInit(Index);
if (isa<InitListExpr>(expr)) {
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getLocStart(), diag::err_init_non_aggr_init_list)
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
<< DeclType << IList->getSourceRange();
hadError = true;
++Index;
++StructuredIndex;
return;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
}
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult Result =
Simplify InitListChecker::CheckReferenceType llvm-svn: 94793
2010-01-29 10:47:33 +08:00
SemaRef.PerformCopyInitialization(Entity, expr->getLocStart(),
SemaRef.Owned(expr));
if (Result.isInvalid())
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
hadError = true;
Simplify InitListChecker::CheckReferenceType llvm-svn: 94793
2010-01-29 10:47:33 +08:00
expr = Result.takeAs<Expr>();
IList->setInit(Index, expr);
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
if (hadError)
++StructuredIndex;
else
UpdateStructuredListElement(StructuredList, StructuredIndex, expr);
++Index;
} else {
Reflow some comments. llvm-svn: 71936
2009-05-16 15:39:55 +08:00
// FIXME: It would be wonderful if we could point at the actual member. In
// general, it would be useful to pass location information down the stack,
// so that we know the location (or decl) of the "current object" being
// initialized.
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(IList->getLocStart(),
Implement and test aggregate initialization in C++. Major changes: - Support initialization of reference members; complain if any reference members are left uninitialized. - Use C++ copy-initialization for initializing each element (falls back to constraint checking in C) - Make sure we diagnose when one tries to provide an initializer list for a non-aggregate. - Don't complain about empty initializers in C++ (they are permitted) - Unrelated but necessary: don't bother trying to convert the decl-specifier-seq to a type when we're dealing with a C++ constructor, destructor, or conversion operator; it results in spurious warnings. llvm-svn: 63431
2009-01-31 06:09:00 +08:00
diag::err_init_reference_member_uninitialized)
<< DeclType
<< IList->getSourceRange();
hadError = true;
++Index;
++StructuredIndex;
return;
}
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckVectorType(const InitializedEntity &Entity,
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
InitListExpr *IList, QualType DeclType,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned &Index,
InitListExpr *StructuredList,
unsigned &StructuredIndex) {
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
if (Index < IList->getNumInits()) {
Change all the Type::getAsFoo() methods to specializations of Type::getAs(). Several of the existing methods were identical to their respective specializations, and so have been removed entirely. Several more 'leaf' optimizations were introduced. The getAsFoo() methods which imposed extra conditions, like getAsObjCInterfacePointerType(), have been left in place. llvm-svn: 82501
2009-09-22 07:43:11 +08:00
const VectorType *VT = DeclType->getAs<VectorType>();
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
unsigned maxElements = VT->getNumElements();
unsigned numEltsInit = 0;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
QualType elementType = VT->getElementType();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
if (!SemaRef.getLangOptions().OpenCL) {
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity ElementEntity =
InitializedEntity::InitializeElement(SemaRef.Context, 0, Entity);
for (unsigned i = 0; i < maxElements; ++i, ++numEltsInit) {
// Don't attempt to go past the end of the init list
if (Index >= IList->getNumInits())
break;
Baby steps towards migrating the InitListChecker over to the new initialization code. Pass an InitializedEntity pointer through to most init checker functions. Right now, it's ignored everywhere except when initializing vectors in C++. llvm-svn: 94325
2010-01-24 03:55:29 +08:00
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
ElementEntity.setElementIndex(Index);
CheckSubElementType(ElementEntity, IList, elementType, Index,
StructuredList, StructuredIndex);
}
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
} else {
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity ElementEntity =
InitializedEntity::InitializeElement(SemaRef.Context, 0, Entity);
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
// OpenCL initializers allows vectors to be constructed from vectors.
for (unsigned i = 0; i < maxElements; ++i) {
// Don't attempt to go past the end of the init list
if (Index >= IList->getNumInits())
break;
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
ElementEntity.setElementIndex(Index);
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
QualType IType = IList->getInit(Index)->getType();
if (!IType->isVectorType()) {
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
CheckSubElementType(ElementEntity, IList, elementType, Index,
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
StructuredList, StructuredIndex);
++numEltsInit;
} else {
Fix an issue with opencl init list checking. llvm-svn: 107824
2010-07-08 06:26:56 +08:00
QualType VecType;
Change all the Type::getAsFoo() methods to specializations of Type::getAs(). Several of the existing methods were identical to their respective specializations, and so have been removed entirely. Several more 'leaf' optimizations were introduced. The getAsFoo() methods which imposed extra conditions, like getAsObjCInterfacePointerType(), have been left in place. llvm-svn: 82501
2009-09-22 07:43:11 +08:00
const VectorType *IVT = IType->getAs<VectorType>();
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
unsigned numIElts = IVT->getNumElements();
Fix an issue with opencl init list checking. llvm-svn: 107824
2010-07-08 06:26:56 +08:00
if (IType->isExtVectorType())
VecType = SemaRef.Context.getExtVectorType(elementType, numIElts);
else
VecType = SemaRef.Context.getVectorType(elementType, numIElts,
IVT->getAltiVecSpecific());
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
CheckSubElementType(ElementEntity, IList, VecType, Index,
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
StructuredList, StructuredIndex);
numEltsInit += numIElts;
}
}
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Fix comment to reflect recent code change. llvm-svn: 101960
2010-04-21 07:21:17 +08:00
// OpenCL requires all elements to be initialized.
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
if (numEltsInit != maxElements)
reapply john's patch, he broke mainline again by changing the test. llvm-svn: 101871
2010-04-20 13:19:10 +08:00
if (SemaRef.getLangOptions().OpenCL)
Take 2 on AltiVec-style vector initializers. Fixes PR4704 problems Addresses Eli's patch feedback re: ugly cast code Updates all postfix operators to remove ParenListExprs. While this is awful, no better solution (say, in the parser) is obvious to me. Better solutions welcome. llvm-svn: 78621
2009-08-11 07:49:36 +08:00
SemaRef.Diag(IList->getSourceRange().getBegin(),
diag::err_vector_incorrect_num_initializers)
<< (numEltsInit < maxElements) << maxElements << numEltsInit;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckArrayType(const InitializedEntity &Entity,
Switch some array initialization over to the new init code. llvm-svn: 94327
2010-01-24 04:13:41 +08:00
InitListExpr *IList, QualType &DeclType,
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
llvm::APSInt elementIndex,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
bool SubobjectIsDesignatorContext,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned &Index,
InitListExpr *StructuredList,
unsigned &StructuredIndex) {
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
// Check for the special-case of initializing an array with a string.
if (Index < IList->getNumInits()) {
handle @encode interactions with array initializers. llvm-svn: 65401
2009-02-25 07:10:27 +08:00
if (Expr *Str = IsStringInit(IList->getInit(Index), DeclType,
SemaRef.Context)) {
CheckStringInit(Str, DeclType, SemaRef);
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// We place the string literal directly into the resulting
// initializer list. This is the only place where the structure
// of the structured initializer list doesn't match exactly,
// because doing so would involve allocating one character
// constant for each string.
make CheckStringLiteralInit a static function in SemaInit.cpp llvm-svn: 65396
2009-02-25 06:41:04 +08:00
UpdateStructuredListElement(StructuredList, StructuredIndex, Str);
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
StructuredList->resizeInits(SemaRef.Context, StructuredIndex);
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
++Index;
return;
}
}
Finally fix PR2189. This makes a fairly invasive but important change to move getAsArrayType into ASTContext instead of being a method on type. This is required because getAsArrayType(const AT), where AT is a typedef for "int[10]" needs to return ArrayType(const int, 10). Fixing this greatly simplifies getArrayDecayedType, which is a good sign. llvm-svn: 54317
2008-08-04 15:31:14 +08:00
if (const VariableArrayType *VAT =
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Context.getAsVariableArrayType(DeclType)) {
Move the error checking for variable-sized objects so we don't double-report errors; fixes PR2362. llvm-svn: 51555
2008-05-25 21:22:35 +08:00
// Check for VLAs; in standard C it would be possible to check this
// earlier, but I don't know where clang accepts VLAs (gcc accepts
// them in all sorts of strange places).
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(VAT->getSizeExpr()->getLocStart(),
remove one more old-style Diag method. llvm-svn: 59589
2008-11-19 13:27:50 +08:00
diag::err_variable_object_no_init)
<< VAT->getSizeExpr()->getSourceRange();
Move the error checking for variable-sized objects so we don't double-report errors; fixes PR2362. llvm-svn: 51555
2008-05-25 21:22:35 +08:00
hadError = true;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++Index;
++StructuredIndex;
Move the error checking for variable-sized objects so we don't double-report errors; fixes PR2362. llvm-svn: 51555
2008-05-25 21:22:35 +08:00
return;
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
// We might know the maximum number of elements in advance.
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
llvm::APSInt maxElements(elementIndex.getBitWidth(),
elementIndex.isUnsigned());
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
bool maxElementsKnown = false;
if (const ConstantArrayType *CAT =
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Context.getAsConstantArrayType(DeclType)) {
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
maxElements = CAT->getSize();
Properly manage the bit-widths of APInts/APSInts in array initialization. Fixes PR clang/3377 llvm-svn: 62851
2009-01-24 00:54:12 +08:00
elementIndex.extOrTrunc(maxElements.getBitWidth());
Make sure all of the isUnsigned flags line up when comparing initializer values, to really really fix PR clang/3377 llvm-svn: 62858
2009-01-24 02:58:42 +08:00
elementIndex.setIsUnsigned(maxElements.isUnsigned());
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
maxElementsKnown = true;
}
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
QualType elementType = SemaRef.Context.getAsArrayType(DeclType)
Finally fix PR2189. This makes a fairly invasive but important change to move getAsArrayType into ASTContext instead of being a method on type. This is required because getAsArrayType(const AT), where AT is a typedef for "int[10]" needs to return ArrayType(const int, 10). Fixing this greatly simplifies getArrayDecayedType, which is a good sign. llvm-svn: 54317
2008-08-04 15:31:14 +08:00
->getElementType();
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
while (Index < IList->getNumInits()) {
Expr *Init = IList->getInit(Index);
if (DesignatedInitExpr *DIE = dyn_cast<DesignatedInitExpr>(Init)) {
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// If we're not the subobject that matches up with the '{' for
// the designator, we shouldn't be handling the
// designator. Return immediately.
if (!SubobjectIsDesignatorContext)
return;
// Handle this designated initializer. elementIndex will be
// updated to be the next array element we'll initialize.
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
if (CheckDesignatedInitializer(Entity, IList, DIE, 0,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
DeclType, 0, &elementIndex, Index,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
StructuredList, StructuredIndex, true,
false)) {
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
hadError = true;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
continue;
}
Properly manage the bit-widths of APInts/APSInts in array initialization. Fixes PR clang/3377 llvm-svn: 62851
2009-01-24 00:54:12 +08:00
if (elementIndex.getBitWidth() > maxElements.getBitWidth())
maxElements.extend(elementIndex.getBitWidth());
else if (elementIndex.getBitWidth() < maxElements.getBitWidth())
elementIndex.extend(maxElements.getBitWidth());
Make sure all of the isUnsigned flags line up when comparing initializer values, to really really fix PR clang/3377 llvm-svn: 62858
2009-01-24 02:58:42 +08:00
elementIndex.setIsUnsigned(maxElements.isUnsigned());
Properly manage the bit-widths of APInts/APSInts in array initialization. Fixes PR clang/3377 llvm-svn: 62851
2009-01-24 00:54:12 +08:00
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// If the array is of incomplete type, keep track of the number of
// elements in the initializer.
if (!maxElementsKnown && elementIndex > maxElements)
maxElements = elementIndex;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
continue;
}
// If we know the maximum number of elements, and we've already
// hit it, stop consuming elements in the initializer list.
if (maxElementsKnown && elementIndex == maxElements)
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
break;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity ElementEntity =
Switch some array initialization over to the new init code. llvm-svn: 94327
2010-01-24 04:13:41 +08:00
InitializedEntity::InitializeElement(SemaRef.Context, StructuredIndex,
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
Entity);
// Check this element.
CheckSubElementType(ElementEntity, IList, elementType, Index,
StructuredList, StructuredIndex);
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
++elementIndex;
// If the array is of incomplete type, keep track of the number of
// elements in the initializer.
if (!maxElementsKnown && elementIndex > maxElements)
maxElements = elementIndex;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Add an assertion so that we don't accidentally build constant arrays of VLAs. llvm-svn: 72587
2009-05-30 04:17:55 +08:00
if (!hadError && DeclType->isIncompleteArrayType()) {
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
// If this is an incomplete array type, the actual type needs to
Support initialization of incomplete array with zero size (as extension). llvm-svn: 54946
2008-08-19 04:28:46 +08:00
// be calculated here.
Make sure all of the isUnsigned flags line up when comparing initializer values, to really really fix PR clang/3377 llvm-svn: 62858
2009-01-24 02:58:42 +08:00
llvm::APSInt Zero(maxElements.getBitWidth(), maxElements.isUnsigned());
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
if (maxElements == Zero) {
Support initialization of incomplete array with zero size (as extension). llvm-svn: 54946
2008-08-19 04:28:46 +08:00
// Sizing an array implicitly to zero is not allowed by ISO C,
// but is supported by GNU.
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IList->getLocStart(),
Support initialization of incomplete array with zero size (as extension). llvm-svn: 54946
2008-08-19 04:28:46 +08:00
diag::ext_typecheck_zero_array_size);
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Support initialization of incomplete array with zero size (as extension). llvm-svn: 54946
2008-08-19 04:28:46 +08:00
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
DeclType = SemaRef.Context.getConstantArrayType(elementType, maxElements,
Support initialization of incomplete array with zero size (as extension). llvm-svn: 54946
2008-08-19 04:28:46 +08:00
ArrayType::Normal, 0);
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
void InitListChecker::CheckStructUnionTypes(const InitializedEntity &Entity,
Use the new init code for member subobjects. llvm-svn: 94329
2010-01-24 04:20:40 +08:00
InitListExpr *IList,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
QualType DeclType,
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
RecordDecl::field_iterator Field,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
bool SubobjectIsDesignatorContext,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned &Index,
InitListExpr *StructuredList,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
unsigned &StructuredIndex,
bool TopLevelObject) {
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
RecordDecl* structDecl = DeclType->getAs<RecordType>()->getDecl();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
// If the record is invalid, some of it's members are invalid. To avoid
// confusion, we forgo checking the intializer for the entire record.
if (structDecl->isInvalidDecl()) {
hadError = true;
return;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
}
Introduce a new expression node, ImplicitValueInitExpr, that represents an implicit value-initialization of a subobject of a particular type. This replaces the (ab)use of CXXZeroValueInitExpr within initializer lists for the "holes" that occur due to the use of C99 designated initializers. The new test case is currently XFAIL'd, because CodeGen's ConstExprEmitter (in lib/CodeGen/CGExprConstant.cpp) needs to be taught to value-initialize when it sees ImplicitValueInitExprs. llvm-svn: 63317
2009-01-30 01:44:32 +08:00
if (DeclType->isUnionType() && IList->getNumInits() == 0) {
// Value-initialize the first named member of the union.
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
RecordDecl *RD = DeclType->getAs<RecordType>()->getDecl();
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
for (RecordDecl::field_iterator FieldEnd = RD->field_end();
Introduce a new expression node, ImplicitValueInitExpr, that represents an implicit value-initialization of a subobject of a particular type. This replaces the (ab)use of CXXZeroValueInitExpr within initializer lists for the "holes" that occur due to the use of C99 designated initializers. The new test case is currently XFAIL'd, because CodeGen's ConstExprEmitter (in lib/CodeGen/CGExprConstant.cpp) needs to be taught to value-initialize when it sees ImplicitValueInitExprs. llvm-svn: 63317
2009-01-30 01:44:32 +08:00
Field != FieldEnd; ++Field) {
if (Field->getDeclName()) {
StructuredList->setInitializedFieldInUnion(*Field);
break;
}
}
return;
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
// If structDecl is a forward declaration, this loop won't do
// anything except look at designated initializers; That's okay,
// because an error should get printed out elsewhere. It might be
// worthwhile to skip over the rest of the initializer, though.
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
RecordDecl *RD = DeclType->getAs<RecordType>()->getDecl();
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
RecordDecl::field_iterator FieldEnd = RD->field_end();
Fix a bug with designated initializers where we were stepping out of a union subobject initialization before checking whether the next initiailizer was actually a designated initializer. This led to spurious "excess elements in union initializer" errors. Thanks to rdivacky for reporting the bug! llvm-svn: 64392
2009-02-13 03:00:39 +08:00
bool InitializedSomething = false;
Implement -Wmissing-field-initializers. Patch by mikem! llvm-svn: 98275
2010-03-12 03:32:38 +08:00
bool CheckForMissingFields = true;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
while (Index < IList->getNumInits()) {
Expr *Init = IList->getInit(Index);
if (DesignatedInitExpr *DIE = dyn_cast<DesignatedInitExpr>(Init)) {
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// If we're not the subobject that matches up with the '{' for
// the designator, we shouldn't be handling the
// designator. Return immediately.
if (!SubobjectIsDesignatorContext)
return;
// Handle this designated initializer. Field will be updated to
// the next field that we'll be initializing.
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
if (CheckDesignatedInitializer(Entity, IList, DIE, 0,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
DeclType, &Field, 0, Index,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
StructuredList, StructuredIndex,
true, TopLevelObject))
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
hadError = true;
Fix a bug with designated initializers where we were stepping out of a union subobject initialization before checking whether the next initiailizer was actually a designated initializer. This led to spurious "excess elements in union initializer" errors. Thanks to rdivacky for reporting the bug! llvm-svn: 64392
2009-02-13 03:00:39 +08:00
InitializedSomething = true;
Implement -Wmissing-field-initializers. Patch by mikem! llvm-svn: 98275
2010-03-12 03:32:38 +08:00
// Disable check for missing fields when designators are used.
// This matches gcc behaviour.
CheckForMissingFields = false;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
continue;
}
if (Field == FieldEnd) {
// We've run out of fields. We're done.
Unifies the name-lookup mechanisms used in various parts of the AST and separates lexical name lookup from qualified name lookup. In particular: * Make DeclContext the central data structure for storing and looking up declarations within existing declarations, e.g., members of structs/unions/classes, enumerators in C++0x enums, members of C++ namespaces, and (later) members of Objective-C interfaces/implementations. DeclContext uses a lazily-constructed data structure optimized for fast lookup (array for small contexts, hash table for larger contexts). * Implement C++ qualified name lookup in terms of lookup into DeclContext. * Implement C++ unqualified name lookup in terms of qualified+unqualified name lookup (since unqualified lookup is not purely lexical in C++!) * Limit the use of the chains of declarations stored in IdentifierInfo to those names declared lexically. * Eliminate CXXFieldDecl, collapsing its behavior into FieldDecl. (FieldDecl is now a ScopedDecl). * Make RecordDecl into a DeclContext and eliminates its Members/NumMembers fields (since one can just iterate through the DeclContext to get the fields). llvm-svn: 60878
2008-12-12 00:49:14 +08:00
break;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
}
Unifies the name-lookup mechanisms used in various parts of the AST and separates lexical name lookup from qualified name lookup. In particular: * Make DeclContext the central data structure for storing and looking up declarations within existing declarations, e.g., members of structs/unions/classes, enumerators in C++0x enums, members of C++ namespaces, and (later) members of Objective-C interfaces/implementations. DeclContext uses a lazily-constructed data structure optimized for fast lookup (array for small contexts, hash table for larger contexts). * Implement C++ qualified name lookup in terms of lookup into DeclContext. * Implement C++ unqualified name lookup in terms of qualified+unqualified name lookup (since unqualified lookup is not purely lexical in C++!) * Limit the use of the chains of declarations stored in IdentifierInfo to those names declared lexically. * Eliminate CXXFieldDecl, collapsing its behavior into FieldDecl. (FieldDecl is now a ScopedDecl). * Make RecordDecl into a DeclContext and eliminates its Members/NumMembers fields (since one can just iterate through the DeclContext to get the fields). llvm-svn: 60878
2008-12-12 00:49:14 +08:00
Fix a bug with designated initializers where we were stepping out of a union subobject initialization before checking whether the next initiailizer was actually a designated initializer. This led to spurious "excess elements in union initializer" errors. Thanks to rdivacky for reporting the bug! llvm-svn: 64392
2009-02-13 03:00:39 +08:00
// We've already initialized a member of a union. We're done.
if (InitializedSomething && DeclType->isUnionType())
break;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
// If we've hit the flexible array member at the end, we're done.
if (Field->getType()->isIncompleteArrayType())
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
break;
Unifies the name-lookup mechanisms used in various parts of the AST and separates lexical name lookup from qualified name lookup. In particular: * Make DeclContext the central data structure for storing and looking up declarations within existing declarations, e.g., members of structs/unions/classes, enumerators in C++0x enums, members of C++ namespaces, and (later) members of Objective-C interfaces/implementations. DeclContext uses a lazily-constructed data structure optimized for fast lookup (array for small contexts, hash table for larger contexts). * Implement C++ qualified name lookup in terms of lookup into DeclContext. * Implement C++ unqualified name lookup in terms of qualified+unqualified name lookup (since unqualified lookup is not purely lexical in C++!) * Limit the use of the chains of declarations stored in IdentifierInfo to those names declared lexically. * Eliminate CXXFieldDecl, collapsing its behavior into FieldDecl. (FieldDecl is now a ScopedDecl). * Make RecordDecl into a DeclContext and eliminates its Members/NumMembers fields (since one can just iterate through the DeclContext to get the fields). llvm-svn: 60878
2008-12-12 00:49:14 +08:00
Clean up designated initialization of unions, so that CodeGen doesn't have to try to guess which member is being initialized. llvm-svn: 63315
2009-01-30 00:53:55 +08:00
if (Field->isUnnamedBitfield()) {
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Don't initialize unnamed bitfields, e.g. "int : 20;"
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
++Field;
Various fixes; solves (almost) all of the test regressions that would be caused by enabling SemaInit. llvm-svn: 51261
2008-05-20 03:16:24 +08:00
continue;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Unifies the name-lookup mechanisms used in various parts of the AST and separates lexical name lookup from qualified name lookup. In particular: * Make DeclContext the central data structure for storing and looking up declarations within existing declarations, e.g., members of structs/unions/classes, enumerators in C++0x enums, members of C++ namespaces, and (later) members of Objective-C interfaces/implementations. DeclContext uses a lazily-constructed data structure optimized for fast lookup (array for small contexts, hash table for larger contexts). * Implement C++ qualified name lookup in terms of lookup into DeclContext. * Implement C++ unqualified name lookup in terms of qualified+unqualified name lookup (since unqualified lookup is not purely lexical in C++!) * Limit the use of the chains of declarations stored in IdentifierInfo to those names declared lexically. * Eliminate CXXFieldDecl, collapsing its behavior into FieldDecl. (FieldDecl is now a ScopedDecl). * Make RecordDecl into a DeclContext and eliminates its Members/NumMembers fields (since one can just iterate through the DeclContext to get the fields). llvm-svn: 60878
2008-12-12 00:49:14 +08:00
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity MemberEntity =
InitializedEntity::InitializeMember(*Field, &Entity);
CheckSubElementType(MemberEntity, IList, Field->getType(), Index,
StructuredList, StructuredIndex);
Fix a bug with designated initializers where we were stepping out of a union subobject initialization before checking whether the next initiailizer was actually a designated initializer. This led to spurious "excess elements in union initializer" errors. Thanks to rdivacky for reporting the bug! llvm-svn: 64392
2009-02-13 03:00:39 +08:00
InitializedSomething = true;
Clean up designated initialization of unions, so that CodeGen doesn't have to try to guess which member is being initialized. llvm-svn: 63315
2009-01-30 00:53:55 +08:00
if (DeclType->isUnionType()) {
// Initialize the first field within the union.
StructuredList->setInitializedFieldInUnion(*Field);
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
++Field;
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Unifies the name-lookup mechanisms used in various parts of the AST and separates lexical name lookup from qualified name lookup. In particular: * Make DeclContext the central data structure for storing and looking up declarations within existing declarations, e.g., members of structs/unions/classes, enumerators in C++0x enums, members of C++ namespaces, and (later) members of Objective-C interfaces/implementations. DeclContext uses a lazily-constructed data structure optimized for fast lookup (array for small contexts, hash table for larger contexts). * Implement C++ qualified name lookup in terms of lookup into DeclContext. * Implement C++ unqualified name lookup in terms of qualified+unqualified name lookup (since unqualified lookup is not purely lexical in C++!) * Limit the use of the chains of declarations stored in IdentifierInfo to those names declared lexically. * Eliminate CXXFieldDecl, collapsing its behavior into FieldDecl. (FieldDecl is now a ScopedDecl). * Make RecordDecl into a DeclContext and eliminates its Members/NumMembers fields (since one can just iterate through the DeclContext to get the fields). llvm-svn: 60878
2008-12-12 00:49:14 +08:00
Implement -Wmissing-field-initializers. Patch by mikem! llvm-svn: 98275
2010-03-12 03:32:38 +08:00
// Emit warnings for missing struct field initializers.
Don't complain about missing field initializers when the initializer list is completely empty, from Lasse Kärkkäinen! Fixes PR7413. llvm-svn: 106320
2010-06-19 05:43:10 +08:00
if (InitializedSomething && CheckForMissingFields && Field != FieldEnd &&
Implement -Wmissing-field-initializers. Patch by mikem! llvm-svn: 98275
2010-03-12 03:32:38 +08:00
!Field->getType()->isIncompleteArrayType() && !DeclType->isUnionType()) {
// It is possible we have one or more unnamed bitfields remaining.
// Find first (if any) named field and emit warning.
for (RecordDecl::field_iterator it = Field, end = RD->field_end();
it != end; ++it) {
if (!it->isUnnamedBitfield()) {
SemaRef.Diag(IList->getSourceRange().getEnd(),
diag::warn_missing_field_initializers) << it->getName();
break;
}
}
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
if (Field == FieldEnd || !Field->getType()->isIncompleteArrayType() ||
Allow flexible array initializers that are not surrounded by braces. We now build the appropriate fully-structured initializer list for such things. Per PR3618, verified that we're getting the right code generation. llvm-svn: 67353
2009-03-20 08:32:56 +08:00
Index >= IList->getNumInits())
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
return;
// Handle GNU flexible array initializers.
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
if (!TopLevelObject &&
Allow flexible array initializers that are not surrounded by braces. We now build the appropriate fully-structured initializer list for such things. Per PR3618, verified that we're getting the right code generation. llvm-svn: 67353
2009-03-20 08:32:56 +08:00
(!isa<InitListExpr>(IList->getInit(Index)) ||
cast<InitListExpr>(IList->getInit(Index))->getNumInits() > 0)) {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(IList->getInit(Index)->getSourceRange().getBegin(),
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
diag::err_flexible_array_init_nonempty)
<< IList->getInit(Index)->getSourceRange().getBegin();
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(Field->getLocation(), diag::note_flexible_array_member)
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
<< *Field;
hadError = true;
Allow flexible array initializers that are not surrounded by braces. We now build the appropriate fully-structured initializer list for such things. Per PR3618, verified that we're getting the right code generation. llvm-svn: 67353
2009-03-20 08:32:56 +08:00
++Index;
return;
} else {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(IList->getInit(Index)->getSourceRange().getBegin(),
Allow flexible array initializers that are not surrounded by braces. We now build the appropriate fully-structured initializer list for such things. Per PR3618, verified that we're getting the right code generation. llvm-svn: 67353
2009-03-20 08:32:56 +08:00
diag::ext_flexible_array_init)
<< IList->getInit(Index)->getSourceRange().getBegin();
SemaRef.Diag(Field->getLocation(), diag::note_flexible_array_member)
<< *Field;
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
}
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity MemberEntity =
InitializedEntity::InitializeMember(*Field, &Entity);
More init work, adding more entity parameters. llvm-svn: 94332
2010-01-24 04:47:59 +08:00
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
if (isa<InitListExpr>(IList->getInit(Index)))
CheckSubElementType(MemberEntity, IList, Field->getType(), Index,
StructuredList, StructuredIndex);
else
CheckImplicitInitList(MemberEntity, IList, Field->getType(), Index,
More init work, adding more entity parameters. llvm-svn: 94332
2010-01-24 04:47:59 +08:00
StructuredList, StructuredIndex);
Replace Sema::CheckInitializerListTypes() with a helper class (InitListChecker) that synthesizes implicit InitListExpr's when appropriate (see InitListExpr comments in Expr.h for more details). I also moved the code to SemaInit.cpp, to help reduce clutter in SemaDecl.cpp. NOTE: This work is incomplete and still fails many tests (as a result, it isn't enabled yet). Nevertheless, I wanted to check it in so I can work on it from home. llvm-svn: 50544
2008-05-02 06:18:59 +08:00
}
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
/// \brief Expand a field designator that refers to a member of an
/// anonymous struct or union into a series of field designators that
/// refers to the field within the appropriate subobject.
///
/// Field/FieldIndex will be updated to point to the (new)
/// currently-designated field.
static void ExpandAnonymousFieldDesignator(Sema &SemaRef,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
DesignatedInitExpr *DIE,
unsigned DesigIdx,
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
FieldDecl *Field,
RecordDecl::field_iterator &FieldIter,
unsigned &FieldIndex) {
typedef DesignatedInitExpr::Designator Designator;
// Build the path from the current object to the member of the
// anonymous struct/union (backwards).
llvm::SmallVector<FieldDecl *, 4> Path;
SemaRef.BuildAnonymousStructUnionMemberPath(Field, Path);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
// Build the replacement designators.
llvm::SmallVector<Designator, 4> Replacements;
for (llvm::SmallVector<FieldDecl *, 4>::reverse_iterator
FI = Path.rbegin(), FIEnd = Path.rend();
FI != FIEnd; ++FI) {
if (FI + 1 == FIEnd)
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Replacements.push_back(Designator((IdentifierInfo *)0,
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
DIE->getDesignator(DesigIdx)->getDotLoc(),
DIE->getDesignator(DesigIdx)->getFieldLoc()));
else
Replacements.push_back(Designator((IdentifierInfo *)0, SourceLocation(),
SourceLocation()));
Replacements.back().setField(*FI);
}
// Expand the current designator into the set of replacement
// designators, so we have a full subobject path down to where the
// member of the anonymous struct/union is actually stored.
Move the allocation of designators in DesignatedInitExpr to the ASTContext. Fixes <rdar://problem/7495428>. llvm-svn: 92867
2010-01-07 07:17:19 +08:00
DIE->ExpandDesignator(SemaRef.Context, DesigIdx, &Replacements[0],
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
&Replacements[0] + Replacements.size());
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
// Update FieldIter/FieldIndex;
RecordDecl *Record = cast<RecordDecl>(Path.back()->getDeclContext());
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
FieldIter = Record->field_begin();
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
FieldIndex = 0;
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
for (RecordDecl::field_iterator FEnd = Record->field_end();
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
FieldIter != FEnd; ++FieldIter) {
if (FieldIter->isUnnamedBitfield())
continue;
if (*FieldIter == Path.back())
return;
++FieldIndex;
}
assert(false && "Unable to find anonymous struct/union field");
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
/// @brief Check the well-formedness of a C99 designated initializer.
///
/// Determines whether the designated initializer @p DIE, which
/// resides at the given @p Index within the initializer list @p
/// IList, is well-formed for a current object of type @p DeclType
/// (C99 6.7.8). The actual subobject that this designator refers to
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
/// within the current subobject is returned in either
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
/// @p NextField or @p NextElementIndex (whichever is appropriate).
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
///
/// @param IList The initializer list in which this designated
/// initializer occurs.
///
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
/// @param DIE The designated initializer expression.
///
/// @param DesigIdx The index of the current designator.
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
///
/// @param DeclType The type of the "current object" (C99 6.7.8p17),
/// into which the designation in @p DIE should refer.
///
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
/// @param NextField If non-NULL and the first designator in @p DIE is
/// a field, this will be set to the field declaration corresponding
/// to the field named by the designator.
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
///
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
/// @param NextElementIndex If non-NULL and the first designator in @p
/// DIE is an array designator or GNU array-range designator, this
/// will be set to the last index initialized by this designator.
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
///
/// @param Index Index into @p IList where the designated initializer
/// @p DIE occurs.
///
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
/// @param StructuredList The initializer list expression that
/// describes all of the subobject initializers in the order they'll
/// actually be initialized.
///
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
/// @returns true if there was an error, false otherwise.
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
bool
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitListChecker::CheckDesignatedInitializer(const InitializedEntity &Entity,
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
InitListExpr *IList,
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
DesignatedInitExpr *DIE,
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
unsigned DesigIdx,
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
QualType &CurrentObjectType,
RecordDecl::field_iterator *NextField,
llvm::APSInt *NextElementIndex,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned &Index,
InitListExpr *StructuredList,
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
unsigned &StructuredIndex,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
bool FinishSubobjectInit,
bool TopLevelObject) {
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
if (DesigIdx == DIE->size()) {
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// Check the actual initialization for the designated object type.
bool prevHadError = hadError;
Eliminate infinite looping in a wacky case with designated initializers. Simplifies (somewhat) the actually checking of the initializer expression following the designators llvm-svn: 63257
2009-01-29 08:39:20 +08:00
// Temporarily remove the designator expression from the
// initializer list that the child calls see, so that we don't try
// to re-process the designator.
unsigned OldIndex = Index;
IList->setInit(OldIndex, DIE->getInit());
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
CheckSubElementType(Entity, IList, CurrentObjectType, Index,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
StructuredList, StructuredIndex);
Eliminate infinite looping in a wacky case with designated initializers. Simplifies (somewhat) the actually checking of the initializer expression following the designators llvm-svn: 63257
2009-01-29 08:39:20 +08:00
// Restore the designated initializer expression in the syntactic
// form of the initializer list.
if (IList->getInit(OldIndex) != DIE->getInit())
DIE->setInit(IList->getInit(OldIndex));
IList->setInit(OldIndex, DIE);
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
return hadError && !prevHadError;
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
bool IsFirstDesignator = (DesigIdx == 0);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
assert((IsFirstDesignator || StructuredList) &&
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
"Need a non-designated initializer list to start from");
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
DesignatedInitExpr::Designator *D = DIE->getDesignator(DesigIdx);
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Determine the structural initializer list that corresponds to the
// current subobject.
StructuredList = IsFirstDesignator? SyntacticToSemantic[IList]
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
: getStructuredSubobjectInit(IList, Index, CurrentObjectType,
Fix PR3509 by providing correct starting locations for initializer lists llvm-svn: 65777
2009-03-02 01:12:46 +08:00
StructuredList, StructuredIndex,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
SourceRange(D->getStartLocation(),
DIE->getSourceRange().getEnd()));
assert(StructuredList && "Expected a structured initializer list");
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
if (D->isFieldDesignator()) {
// C99 6.7.8p7:
//
// If a designator has the form
//
// . identifier
//
// then the current object (defined below) shall have
// structure or union type and the identifier shall be the
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
// name of a member of that type.
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
const RecordType *RT = CurrentObjectType->getAs<RecordType>();
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
if (!RT) {
SourceLocation Loc = D->getDotLoc();
if (Loc.isInvalid())
Loc = D->getFieldLoc();
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(Loc, diag::err_field_designator_non_aggr)
<< SemaRef.getLangOptions().CPlusPlus << CurrentObjectType;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
++Index;
return true;
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Note: we perform a linear search of the fields here, despite
// the fact that we have a faster lookup method, because we always
// need to compute the field's index.
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
FieldDecl *KnownField = D->getField();
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
IdentifierInfo *FieldName = D->getFieldName();
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
unsigned FieldIndex = 0;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
RecordDecl::field_iterator
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
Field = RT->getDecl()->field_begin(),
FieldEnd = RT->getDecl()->field_end();
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
for (; Field != FieldEnd; ++Field) {
if (Field->isUnnamedBitfield())
continue;
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
if (KnownField == *Field || Field->getIdentifier() == FieldName)
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
break;
++FieldIndex;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
if (Field == FieldEnd) {
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
// There was no normal field in the struct with the designated
// name. Perform another lookup for this name, which may find
// something that we can't designate (e.g., a member function),
// may find nothing, or may find a member of an anonymous
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
// struct/union.
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
DeclContext::lookup_result Lookup = RT->getDecl()->lookup(FieldName);
Typo correction for C99 designated field initializers, e.g., test/FixIt/typo.c:19:4: error: field designator 'bunds' does not refer to any field in type 'struct Window'; did you mean 'bounds'? .bunds. ^~~~~ bounds llvm-svn: 92376
2010-01-01 08:03:05 +08:00
FieldDecl *ReplacementField = 0;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
if (Lookup.first == Lookup.second) {
Typo correction for C99 designated field initializers, e.g., test/FixIt/typo.c:19:4: error: field designator 'bunds' does not refer to any field in type 'struct Window'; did you mean 'bounds'? .bunds. ^~~~~ bounds llvm-svn: 92376
2010-01-01 08:03:05 +08:00
// Name lookup didn't find anything. Determine whether this
// was a typo for another field name.
LookupResult R(SemaRef, FieldName, D->getFieldLoc(),
Sema::LookupMemberName);
Teach typo correction about various language keywords. We can't generally recover from typos in keywords (since we would effectively have to mangle the token stream). However, there are still benefits to typo-correcting with keywords: - We don't make stupid suggestions when the user typed something that is similar to a keyword. - We can suggest the keyword in a diagnostic (did you mean "static_cast"?), even if we can't recover and therefore don't have a fix-it. llvm-svn: 101274
2010-04-15 04:04:41 +08:00
if (SemaRef.CorrectTypo(R, /*Scope=*/0, /*SS=*/0, RT->getDecl(), false,
Sema::CTC_NoKeywords) &&
Typo correction for C99 designated field initializers, e.g., test/FixIt/typo.c:19:4: error: field designator 'bunds' does not refer to any field in type 'struct Window'; did you mean 'bounds'? .bunds. ^~~~~ bounds llvm-svn: 92376
2010-01-01 08:03:05 +08:00
(ReplacementField = R.getAsSingle<FieldDecl>()) &&
Rename DeclContext::getLookupContext to getRedeclContext and change its semantics slightly. No functionality change in the absence of inline namespaces. Also, change a few places where inline namespaces actually make a difference to be prepared for them. llvm-svn: 112563
2010-08-31 08:36:30 +08:00
ReplacementField->getDeclContext()->getRedeclContext()
Typo correction for C99 designated field initializers, e.g., test/FixIt/typo.c:19:4: error: field designator 'bunds' does not refer to any field in type 'struct Window'; did you mean 'bounds'? .bunds. ^~~~~ bounds llvm-svn: 92376
2010-01-01 08:03:05 +08:00
->Equals(RT->getDecl())) {
SemaRef.Diag(D->getFieldLoc(),
diag::err_field_designator_unknown_suggest)
<< FieldName << CurrentObjectType << R.getLookupName()
Reinstate my CodeModificationHint -> FixItHint renaming patch, without the C-only "optimization". llvm-svn: 100022
2010-04-01 01:46:05 +08:00
<< FixItHint::CreateReplacement(D->getFieldLoc(),
R.getLookupName().getAsString());
Whenever we emit a typo-correction diagnostic, also emit a note pointing to the declaration that we found that has that name (if it is unique). llvm-svn: 92877
2010-01-07 08:17:44 +08:00
SemaRef.Diag(ReplacementField->getLocation(),
diag::note_previous_decl)
<< ReplacementField->getDeclName();
Typo correction for C99 designated field initializers, e.g., test/FixIt/typo.c:19:4: error: field designator 'bunds' does not refer to any field in type 'struct Window'; did you mean 'bounds'? .bunds. ^~~~~ bounds llvm-svn: 92376
2010-01-01 08:03:05 +08:00
} else {
SemaRef.Diag(D->getFieldLoc(), diag::err_field_designator_unknown)
<< FieldName << CurrentObjectType;
++Index;
return true;
}
} else if (!KnownField) {
// Determine whether we found a field at all.
ReplacementField = dyn_cast<FieldDecl>(*Lookup.first);
}
if (!ReplacementField) {
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Name lookup found something, but it wasn't a field.
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(D->getFieldLoc(), diag::err_field_designator_nonfield)
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
<< FieldName;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag((*Lookup.first)->getLocation(),
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
diag::note_field_designator_found);
Attempt to fix a read-after-free running test/Sema/designated-initializers.c. Douglas, can you check that this is doing the right thing? llvm-svn: 69298
2009-04-17 01:49:48 +08:00
++Index;
return true;
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
}
Typo correction for C99 designated field initializers, e.g., test/FixIt/typo.c:19:4: error: field designator 'bunds' does not refer to any field in type 'struct Window'; did you mean 'bounds'? .bunds. ^~~~~ bounds llvm-svn: 92376
2010-01-01 08:03:05 +08:00
if (!KnownField &&
cast<RecordDecl>((ReplacementField)->getDeclContext())
->isAnonymousStructOrUnion()) {
// Handle an field designator that refers to a member of an
// anonymous struct or union.
ExpandAnonymousFieldDesignator(SemaRef, DIE, DesigIdx,
ReplacementField,
Field, FieldIndex);
D = DIE->getDesignator(DesigIdx);
} else if (!KnownField) {
// The replacement field comes from typo correction; find it
// in the list of fields.
FieldIndex = 0;
Field = RT->getDecl()->field_begin();
for (; Field != FieldEnd; ++Field) {
if (Field->isUnnamedBitfield())
continue;
if (ReplacementField == *Field ||
Field->getIdentifier() == ReplacementField->getIdentifier())
break;
++FieldIndex;
}
}
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
} else if (!KnownField &&
cast<RecordDecl>((*Field)->getDeclContext())
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
->isAnonymousStructOrUnion()) {
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
ExpandAnonymousFieldDesignator(SemaRef, DIE, DesigIdx, *Field,
Field, FieldIndex);
D = DIE->getDesignator(DesigIdx);
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// All of the fields of a union are located at the same place in
// the initializer list.
Clean up designated initialization of unions, so that CodeGen doesn't have to try to guess which member is being initialized. llvm-svn: 63315
2009-01-30 00:53:55 +08:00
if (RT->getDecl()->isUnion()) {
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
FieldIndex = 0;
Clean up designated initialization of unions, so that CodeGen doesn't have to try to guess which member is being initialized. llvm-svn: 63315
2009-01-30 00:53:55 +08:00
StructuredList->setInitializedFieldInUnion(*Field);
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// Update the designator with the field declaration.
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
D->setField(*Field);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Make sure that our non-designated initializer list has space
// for a subobject corresponding to this field.
if (FieldIndex >= StructuredList->getNumInits())
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
StructuredList->resizeInits(SemaRef.Context, FieldIndex + 1);
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
// This designator names a flexible array member.
if (Field->getType()->isIncompleteArrayType()) {
bool Invalid = false;
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
if ((DesigIdx + 1) != DIE->size()) {
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
// We can't designate an object within the flexible array
// member (because GCC doesn't allow it).
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
DesignatedInitExpr::Designator *NextD
Switch designated-initializer checking from using designator iterators to using designator indices. No functionality change. llvm-svn: 69147
2009-04-15 12:56:10 +08:00
= DIE->getDesignator(DesigIdx + 1);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(NextD->getStartLocation(),
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
diag::err_designator_into_flexible_array_member)
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
<< SourceRange(NextD->getStartLocation(),
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
DIE->getSourceRange().getEnd());
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(Field->getLocation(), diag::note_flexible_array_member)
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
<< *Field;
Invalid = true;
}
if (!hadError && !isa<InitListExpr>(DIE->getInit())) {
// The initializer is not an initializer list.
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(DIE->getInit()->getSourceRange().getBegin(),
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
diag::err_flexible_array_init_needs_braces)
<< DIE->getInit()->getSourceRange();
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(Field->getLocation(), diag::note_flexible_array_member)
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
<< *Field;
Invalid = true;
}
// Handle GNU flexible array initializers.
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
if (!Invalid && !TopLevelObject &&
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
cast<InitListExpr>(DIE->getInit())->getNumInits() > 0) {
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(DIE->getSourceRange().getBegin(),
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
diag::err_flexible_array_init_nonempty)
<< DIE->getSourceRange().getBegin();
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(Field->getLocation(), diag::note_flexible_array_member)
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
<< *Field;
Invalid = true;
}
if (Invalid) {
++Index;
return true;
}
// Initialize the array.
bool prevHadError = hadError;
unsigned newStructuredIndex = FieldIndex;
unsigned OldIndex = Index;
IList->setInit(Index, DIE->getInit());
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity MemberEntity =
InitializedEntity::InitializeMember(*Field, &Entity);
CheckSubElementType(MemberEntity, IList, Field->getType(), Index,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
StructuredList, newStructuredIndex);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
IList->setInit(OldIndex, DIE);
if (hadError && !prevHadError) {
++Field;
++FieldIndex;
if (NextField)
*NextField = Field;
StructuredIndex = FieldIndex;
return true;
}
} else {
// Recurse to check later designated subobjects.
QualType FieldType = (*Field)->getType();
unsigned newStructuredIndex = FieldIndex;
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
InitializedEntity MemberEntity =
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity::InitializeMember(*Field, &Entity);
if (CheckDesignatedInitializer(MemberEntity, IList, DIE, DesigIdx + 1,
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
FieldType, 0, 0, Index,
StructuredList, newStructuredIndex,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
true, false))
return true;
}
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// Find the position of the next field to be initialized in this
// subobject.
++Field;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++FieldIndex;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// If this the first designator, our caller will continue checking
// the rest of this struct/class/union subobject.
if (IsFirstDesignator) {
if (NextField)
*NextField = Field;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
StructuredIndex = FieldIndex;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
return false;
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
if (!FinishSubobjectInit)
return false;
Implement support for designated initializers that refer to members of anonymous structs or unions. Fixes PR3778. llvm-svn: 69153
2009-04-15 14:41:24 +08:00
// We've already initialized something in the union; we're done.
if (RT->getDecl()->isUnion())
return hadError;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// Check the remaining fields within this class/struct/union subobject.
bool prevHadError = hadError;
Use the new init code for member subobjects. llvm-svn: 94329
2010-01-24 04:20:40 +08:00
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
CheckStructUnionTypes(Entity, IList, CurrentObjectType, Field, false, Index,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
StructuredList, FieldIndex);
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
return hadError && !prevHadError;
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// C99 6.7.8p6:
//
// If a designator has the form
//
// [ constant-expression ]
//
// then the current object (defined below) shall have array
// type and the expression shall be an integer constant
// expression. If the array is of unknown size, any
// nonnegative value is valid.
//
// Additionally, cope with the GNU extension that permits
// designators of the form
//
// [ constant-expression ... constant-expression ]
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
const ArrayType *AT = SemaRef.Context.getAsArrayType(CurrentObjectType);
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
if (!AT) {
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(D->getLBracketLoc(), diag::err_array_designator_non_array)
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
<< CurrentObjectType;
++Index;
return true;
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
Expr *IndexExpr = 0;
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
llvm::APSInt DesignatedStartIndex, DesignatedEndIndex;
if (D->isArrayDesignator()) {
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
IndexExpr = DIE->getArrayIndex(*D);
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
DesignatedStartIndex = IndexExpr->EvaluateAsInt(SemaRef.Context);
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
DesignatedEndIndex = DesignatedStartIndex;
} else {
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
assert(D->isArrayRangeDesignator() && "Need array-range designator");
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
DesignatedStartIndex =
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
DIE->getArrayRangeStart(*D)->EvaluateAsInt(SemaRef.Context);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
DesignatedEndIndex =
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
DIE->getArrayRangeEnd(*D)->EvaluateAsInt(SemaRef.Context);
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
IndexExpr = DIE->getArrayRangeEnd(*D);
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
if (DesignatedStartIndex.getZExtValue() !=DesignatedEndIndex.getZExtValue())
Make CodeGen produce an error if we come across a non-constant initializer list that involves the GNU array-range designator extension llvm-svn: 63327
2009-01-30 03:42:23 +08:00
FullyStructuredList->sawArrayRangeDesignator();
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
if (isa<ConstantArrayType>(AT)) {
llvm::APSInt MaxElements(cast<ConstantArrayType>(AT)->getSize(), false);
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
DesignatedStartIndex.extOrTrunc(MaxElements.getBitWidth());
DesignatedStartIndex.setIsUnsigned(MaxElements.isUnsigned());
DesignatedEndIndex.extOrTrunc(MaxElements.getBitWidth());
DesignatedEndIndex.setIsUnsigned(MaxElements.isUnsigned());
if (DesignatedEndIndex >= MaxElements) {
make SemaRef be a reference to sema, not a pointer. llvm-svn: 65399
2009-02-25 06:50:46 +08:00
SemaRef.Diag(IndexExpr->getSourceRange().getBegin(),
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
diag::err_array_designator_too_large)
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
<< DesignatedEndIndex.toString(10) << MaxElements.toString(10)
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
<< IndexExpr->getSourceRange();
++Index;
return true;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
}
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
} else {
// Make sure the bit-widths and signedness match.
if (DesignatedStartIndex.getBitWidth() > DesignatedEndIndex.getBitWidth())
DesignatedEndIndex.extend(DesignatedStartIndex.getBitWidth());
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
else if (DesignatedStartIndex.getBitWidth() <
DesignatedEndIndex.getBitWidth())
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
DesignatedStartIndex.extend(DesignatedEndIndex.getBitWidth());
DesignatedStartIndex.setIsUnsigned(true);
DesignatedEndIndex.setIsUnsigned(true);
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Make sure that our non-designated initializer list has space
// for a subobject corresponding to this array element.
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
if (DesignatedEndIndex.getZExtValue() >= StructuredList->getNumInits())
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
StructuredList->resizeInits(SemaRef.Context,
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
DesignatedEndIndex.getZExtValue() + 1);
// Repeatedly perform subobject initializations in the range
// [DesignatedStartIndex, DesignatedEndIndex].
// Move to the next designator
unsigned ElementIndex = DesignatedStartIndex.getZExtValue();
unsigned OldIndex = Index;
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
InitializedEntity ElementEntity =
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
InitializedEntity::InitializeElement(SemaRef.Context, 0, Entity);
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
while (DesignatedStartIndex <= DesignatedEndIndex) {
// Recurse to check later designated subobjects.
QualType ElementType = AT->getElementType();
Index = OldIndex;
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
ElementEntity.setElementIndex(ElementIndex);
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
if (CheckDesignatedInitializer(ElementEntity, IList, DIE, DesigIdx + 1,
Start passing InitializedEntity to CheckDesignatedInitializer. llvm-svn: 94334
2010-01-24 06:49:02 +08:00
ElementType, 0, 0, Index,
StructuredList, ElementIndex,
Implement semantic analysis for the GNU flexible array initialization extension. The interaction with designated initializers is a bit... interesting... but we follow GNU's lead and don't permit too much crazy code in this area. Also, make the "excess initializers" error message a bit more informative. Addresses PR2561: http://llvm.org/bugs/show_bug.cgi?id=2561 llvm-svn: 63785
2009-02-05 06:46:25 +08:00
(DesignatedStartIndex == DesignatedEndIndex),
false))
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
return true;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
// Move to the next index in the array that we'll be initializing.
++DesignatedStartIndex;
ElementIndex = DesignatedStartIndex.getZExtValue();
}
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// If this the first designator, our caller will continue checking
// the rest of this array subobject.
if (IsFirstDesignator) {
if (NextElementIndex)
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
*NextElementIndex = DesignatedStartIndex;
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
StructuredIndex = ElementIndex;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
return false;
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Improvements to code-generation and semantic analysis of designated initializers. - We now initialize unions properly when a member other than the first is named by a designated initializer. - We now provide proper semantic analysis and code generation for GNU array-range designators *except* that side effects will occur more than once. We warn about this. llvm-svn: 63253
2009-01-29 07:36:17 +08:00
if (!FinishSubobjectInit)
return false;
Reimplement the handling of the "current object" in designator initializers, so that we are within the appropriate subobject after we've processed a multi-designator designation. We're matching GCC and EDG's behavior on all examples I've found thus far. *Huge* thanks to Eli Friedman for pointing out my fundamental misunderstanding of "current object" in the C99 spec. llvm-svn: 62812
2009-01-23 07:26:18 +08:00
// Check the remaining elements within this array subobject.
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
bool prevHadError = hadError;
Change all InitializedEntity pointers over to be references. llvm-svn: 94335
2010-01-24 07:23:01 +08:00
CheckArrayType(Entity, IList, CurrentObjectType, DesignatedStartIndex,
Switch some array initialization over to the new init code. llvm-svn: 94327
2010-01-24 04:13:41 +08:00
/*SubobjectIsDesignatorContext=*/false, Index,
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
StructuredList, ElementIndex);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
return hadError && !prevHadError;
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
}
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Get the structured initializer list for a subobject of type
// @p CurrentObjectType.
InitListExpr *
InitListChecker::getStructuredSubobjectInit(InitListExpr *IList, unsigned Index,
QualType CurrentObjectType,
InitListExpr *StructuredList,
unsigned StructuredIndex,
SourceRange InitRange) {
Expr *ExistingInit = 0;
if (!StructuredList)
ExistingInit = SyntacticToSemantic[IList];
else if (StructuredIndex < StructuredList->getNumInits())
ExistingInit = StructuredList->getInit(StructuredIndex);
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
if (InitListExpr *Result = dyn_cast_or_null<InitListExpr>(ExistingInit))
return Result;
if (ExistingInit) {
// We are creating an initializer list that initializes the
// subobjects of the current object, but there was already an
// initialization that completely initialized the current
// subobject, e.g., by a compound literal:
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
//
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// struct X { int a, b; };
// struct X xs[] = { [0] = (struct X) { 1, 2 }, [0].b = 3 };
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
//
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Here, xs[0].a == 0 and xs[0].b == 3, since the second,
// designated initializer re-initializes the whole
// subobject [0], overwriting previous initializers.
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(InitRange.getBegin(),
Fix PR3509 by providing correct starting locations for initializer lists llvm-svn: 65777
2009-03-02 01:12:46 +08:00
diag::warn_subobject_initializer_overrides)
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
<< InitRange;
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(ExistingInit->getSourceRange().getBegin(),
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
diag::note_previous_initializer)
Remove Expr::hasSideEffects. It doesn't work anyway llvm-svn: 63254
2009-01-29 07:43:32 +08:00
<< /*FIXME:has side effects=*/0
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
<< ExistingInit->getSourceRange();
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
InitListExpr *Result
Use ASTVector instead of std::vector for the Exprs in InitListExpr. Performance measurements of '-fsyntax-only' on combine.c (403.gcc) shows no real performance change, but now the vector isn't leaked. llvm-svn: 101195
2010-04-14 07:39:13 +08:00
= new (SemaRef.Context) InitListExpr(SemaRef.Context,
InitRange.getBegin(), 0, 0,
Revert: "Change InitListExpr to allocate the array for holding references" This was causing buildbot breakage. This reverts commit d46e952cc8cb8d9eed8657d9a0b267910a0f745a. llvm-svn: 96652
2010-02-19 09:50:18 +08:00
InitRange.getEnd());
Fix PR3509 by providing correct starting locations for initializer lists llvm-svn: 65777
2009-03-02 01:12:46 +08:00
Whenever we're creating an expression that is typically an rvalue (e.g., a call, cast, etc.), immediately adjust the expression's type to strip cv-qualifiers off of all non-class types (in C++) or all types (in C). This effectively extends my previous fix for PR7463, which was restricted to calls, to other kinds of expressions within similar characteristics. I've audited every use of getNonReferenceType() in the code base, switching to the newly-renamed getNonLValueExprType() where necessary. Big thanks to Eli for pointing out just how incomplete my original fix for PR7463 actually was. We've been handling cv-qualifiers on rvalues wrong for a very, very long time. Fixes PR7463. llvm-svn: 108253
2010-07-14 02:40:04 +08:00
Result->setType(CurrentObjectType.getNonLValueExprType(SemaRef.Context));
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
// Pre-allocate storage for the structured initializer list.
unsigned NumElements = 0;
Fix a thinko in the pre-allocation strategy for structured initializer lists. The code wasn't accounting for the distinction between the top-level call to getStructuredSubobjectInit and later calls that occur deeper in the hierarchy. This problem manifested itself as over-allocation in cases where we have large arrays of small structures (<rdar://problem/6707362>). llvm-svn: 67452
2009-03-22 02:13:52 +08:00
unsigned NumInits = 0;
if (!StructuredList)
NumInits = IList->getNumInits();
else if (Index < IList->getNumInits()) {
if (InitListExpr *SubList = dyn_cast<InitListExpr>(IList->getInit(Index)))
NumInits = SubList->getNumInits();
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
if (const ArrayType *AType
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
= SemaRef.Context.getAsArrayType(CurrentObjectType)) {
if (const ConstantArrayType *CAType = dyn_cast<ConstantArrayType>(AType)) {
NumElements = CAType->getSize().getZExtValue();
// Simple heuristic so that we don't allocate a very large
// initializer with many empty entries at the end.
Fix a thinko in the pre-allocation strategy for structured initializer lists. The code wasn't accounting for the distinction between the top-level call to getStructuredSubobjectInit and later calls that occur deeper in the hierarchy. This problem manifested itself as over-allocation in cases where we have large arrays of small structures (<rdar://problem/6707362>). llvm-svn: 67452
2009-03-22 02:13:52 +08:00
if (NumInits && NumElements > NumInits)
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
NumElements = 0;
}
Change all the Type::getAsFoo() methods to specializations of Type::getAs(). Several of the existing methods were identical to their respective specializations, and so have been removed entirely. Several more 'leaf' optimizations were introduced. The getAsFoo() methods which imposed extra conditions, like getAsObjCInterfacePointerType(), have been left in place. llvm-svn: 82501
2009-09-22 07:43:11 +08:00
} else if (const VectorType *VType = CurrentObjectType->getAs<VectorType>())
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
NumElements = VType->getNumElements();
Change uses of: Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsRecordType() -> Type::getAs<RecordType>() Type::getAsPointerType() -> Type::getAs<PointerType>() Type::getAsBlockPointerType() -> Type::getAs<BlockPointerType>() Type::getAsLValueReferenceType() -> Type::getAs<LValueReferenceType>() Type::getAsRValueReferenceType() -> Type::getAs<RValueReferenceType>() Type::getAsMemberPointerType() -> Type::getAs<MemberPointerType>() Type::getAsReferenceType() -> Type::getAs<ReferenceType>() Type::getAsTagType() -> Type::getAs<TagType>() And remove Type::getAsReferenceType(), etc. This change is similar to one I made a couple weeks ago, but that was partly reverted pending some additional design discussion. With Doug's pending smart pointer changes for Types, it seemed natural to take this approach. llvm-svn: 77510
2009-07-30 05:53:49 +08:00
else if (const RecordType *RType = CurrentObjectType->getAs<RecordType>()) {
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
RecordDecl *RDecl = RType->getDecl();
if (RDecl->isUnion())
NumElements = 1;
else
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
NumElements = std::distance(RDecl->field_begin(),
De-ASTContext-ify DeclContext. Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating". Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit. llvm-svn: 74506
2009-06-30 10:36:12 +08:00
RDecl->field_end());
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
}
Fix a thinko in the pre-allocation strategy for structured initializer lists. The code wasn't accounting for the distinction between the top-level call to getStructuredSubobjectInit and later calls that occur deeper in the hierarchy. This problem manifested itself as over-allocation in cases where we have large arrays of small structures (<rdar://problem/6707362>). llvm-svn: 67452
2009-03-22 02:13:52 +08:00
if (NumElements < NumInits)
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
NumElements = IList->getNumInits();
Use ASTVector instead of std::vector for the Exprs in InitListExpr. Performance measurements of '-fsyntax-only' on combine.c (403.gcc) shows no real performance change, but now the vector isn't leaked. llvm-svn: 101195
2010-04-14 07:39:13 +08:00
Result->reserveInits(SemaRef.Context, NumElements);
When building the structured initializer list, pre-allocate storage in its vectors based on the subobject type we're initializing and the (unstructured) initializer list. This eliminates some malloc thrashing when parsing initializers (from 117 vector reallocations down to 0 when parsing Cocoa.h). We can't always pre-allocate the right amount of storage, since designated initializers can cause us to initialize in non-predictable patterns. llvm-svn: 67421
2009-03-21 07:58:33 +08:00
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// Link this new initializer list into the structured initializer
// lists.
if (StructuredList)
Use ASTVector instead of std::vector for the Exprs in InitListExpr. Performance measurements of '-fsyntax-only' on combine.c (403.gcc) shows no real performance change, but now the vector isn't leaked. llvm-svn: 101195
2010-04-14 07:39:13 +08:00
StructuredList->updateInit(SemaRef.Context, StructuredIndex, Result);
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
else {
Result->setSyntacticForm(IList);
SyntacticToSemantic[IList] = Result;
}
return Result;
}
/// Update the initializer at index @p StructuredIndex within the
/// structured initializer list to the value @p expr.
void InitListChecker::UpdateStructuredListElement(InitListExpr *StructuredList,
unsigned &StructuredIndex,
Expr *expr) {
// No structured initializer list to update
if (!StructuredList)
return;
Use ASTVector instead of std::vector for the Exprs in InitListExpr. Performance measurements of '-fsyntax-only' on combine.c (403.gcc) shows no real performance change, but now the vector isn't leaked. llvm-svn: 101195
2010-04-14 07:39:13 +08:00
if (Expr *PrevInit = StructuredList->updateInit(SemaRef.Context,
StructuredIndex, expr)) {
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
// This initializer overwrites a previous initializer. Warn.
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(expr->getSourceRange().getBegin(),
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
diag::warn_initializer_overrides)
<< expr->getSourceRange();
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
SemaRef.Diag(PrevInit->getSourceRange().getBegin(),
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
diag::note_previous_initializer)
Remove Expr::hasSideEffects. It doesn't work anyway llvm-svn: 63254
2009-01-29 07:43:32 +08:00
<< /*FIXME:has side effects=*/0
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
<< PrevInit->getSourceRange();
}
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Code generation support for C99 designated initializers. The approach I've taken in this patch is relatively straightforward, although the code itself is non-trivial. Essentially, as we process an initializer list we build up a fully-explicit representation of the initializer list, where each of the subobject initializations occurs in order. Designators serve to "fill in" subobject initializations in a non-linear way. The fully-explicit representation makes initializer lists (both with and without designators) easy to grok for codegen and later semantic analyses. We keep the syntactic form of the initializer list linked into the AST for those clients interested in exactly what the user wrote. Known limitations: - Designating a member of a union that isn't the first member may result in bogus initialization (we warn about this) - GNU array-range designators are not supported (we warn about this) llvm-svn: 63242
2009-01-29 05:54:33 +08:00
++StructuredIndex;
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
/// Check that the given Index expression is a valid array designator
/// value. This is essentailly just a wrapper around
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
/// VerifyIntegerConstantExpression that also checks for negative values
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
/// and produces a reasonable diagnostic if there is a
/// failure. Returns true if there was an error, false otherwise. If
/// everything went okay, Value will receive the value of the constant
/// expression.
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
static bool
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
CheckArrayDesignatorExpr(Sema &S, Expr *Index, llvm::APSInt &Value) {
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
SourceLocation Loc = Index->getSourceRange().getBegin();
// Make sure this is an integer constant expression.
fix PR4073 by making designated initializer checking code use VerifyIntegerConstantExpression instead of isIntegerConstantExpr. This makes it ext-warn but tolerate things that fold to a constant but that are not valid i-c-e's. There must be a bug in the i-c-e computation though, because it doesn't catch this case even with pedantic. This also switches the later code to use EvaluateAsInt which is simpler and handles everything that evaluate does. llvm-svn: 70081
2009-04-26 05:59:05 +08:00
if (S.VerifyIntegerConstantExpression(Index, &Value))
return true;
if (Value.isSigned() && Value.isNegative())
return S.Diag(Loc, diag::err_array_designator_negative)
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
<< Value.toString(10) << Index->getSourceRange();
Hopefully the last of the APSInt signedness issues with initializers. Fixes PR clang/3378 llvm-svn: 62876
2009-01-24 05:04:18 +08:00
Value.setIsUnsigned(true);
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
return false;
}
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult Sema::ActOnDesignatedInitializer(Designation &Desig,
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
SourceLocation Loc,
Make our diagnostics about the obsolete GNU designated-initializer syntax into extension warnings, and provide code-modification hints showing how to fix the problem. llvm-svn: 67885
2009-03-28 08:41:23 +08:00
bool GNUSyntax,
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult Init) {
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
typedef DesignatedInitExpr::Designator ASTDesignator;
bool Invalid = false;
llvm::SmallVector<ASTDesignator, 32> Designators;
llvm::SmallVector<Expr *, 32> InitExpressions;
// Build designators and check array designator expressions.
for (unsigned Idx = 0; Idx < Desig.getNumDesignators(); ++Idx) {
const Designator &D = Desig.getDesignator(Idx);
switch (D.getKind()) {
case Designator::FieldDesignator:
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
Designators.push_back(ASTDesignator(D.getField(), D.getDotLoc(),
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
D.getFieldLoc()));
break;
case Designator::ArrayDesignator: {
Expr *Index = static_cast<Expr *>(D.getArrayIndex());
llvm::APSInt IndexValue;
Template instantiation for C99 designated initializers, because we can. Also, delay semantic analysis of initialization for value-dependent as well as type-dependent expressions, since we can't always properly type-check a value-dependent expression. llvm-svn: 72233
2009-05-22 07:17:49 +08:00
if (!Index->isTypeDependent() &&
!Index->isValueDependent() &&
CheckArrayDesignatorExpr(*this, Index, IndexValue))
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Invalid = true;
else {
Designators.push_back(ASTDesignator(InitExpressions.size(),
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
D.getLBracketLoc(),
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
D.getRBracketLoc()));
InitExpressions.push_back(Index);
}
break;
}
case Designator::ArrayRangeDesignator: {
Expr *StartIndex = static_cast<Expr *>(D.getArrayRangeStart());
Expr *EndIndex = static_cast<Expr *>(D.getArrayRangeEnd());
llvm::APSInt StartValue;
llvm::APSInt EndValue;
Template instantiation for C99 designated initializers, because we can. Also, delay semantic analysis of initialization for value-dependent as well as type-dependent expressions, since we can't always properly type-check a value-dependent expression. llvm-svn: 72233
2009-05-22 07:17:49 +08:00
bool StartDependent = StartIndex->isTypeDependent() ||
StartIndex->isValueDependent();
bool EndDependent = EndIndex->isTypeDependent() ||
EndIndex->isValueDependent();
if ((!StartDependent &&
CheckArrayDesignatorExpr(*this, StartIndex, StartValue)) ||
(!EndDependent &&
CheckArrayDesignatorExpr(*this, EndIndex, EndValue)))
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
Invalid = true;
More APSInt appeasement llvm-svn: 62884
2009-01-24 06:22:29 +08:00
else {
// Make sure we're comparing values with the same bit width.
Template instantiation for C99 designated initializers, because we can. Also, delay semantic analysis of initialization for value-dependent as well as type-dependent expressions, since we can't always properly type-check a value-dependent expression. llvm-svn: 72233
2009-05-22 07:17:49 +08:00
if (StartDependent || EndDependent) {
// Nothing to compute.
} else if (StartValue.getBitWidth() > EndValue.getBitWidth())
More APSInt appeasement llvm-svn: 62884
2009-01-24 06:22:29 +08:00
EndValue.extend(StartValue.getBitWidth());
else if (StartValue.getBitWidth() < EndValue.getBitWidth())
StartValue.extend(EndValue.getBitWidth());
Template instantiation for GNU array-range designators. llvm-svn: 72234
2009-05-22 07:30:39 +08:00
if (!StartDependent && !EndDependent && EndValue < StartValue) {
More APSInt appeasement llvm-svn: 62884
2009-01-24 06:22:29 +08:00
Diag(D.getEllipsisLoc(), diag::err_array_designator_empty_range)
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
<< StartValue.toString(10) << EndValue.toString(10)
More APSInt appeasement llvm-svn: 62884
2009-01-24 06:22:29 +08:00
<< StartIndex->getSourceRange() << EndIndex->getSourceRange();
Invalid = true;
} else {
Designators.push_back(ASTDesignator(InitExpressions.size(),
Remove tabs, and whitespace cleanups. llvm-svn: 81346
2009-09-09 23:08:12 +08:00
D.getLBracketLoc(),
More APSInt appeasement llvm-svn: 62884
2009-01-24 06:22:29 +08:00
D.getEllipsisLoc(),
D.getRBracketLoc()));
InitExpressions.push_back(StartIndex);
InitExpressions.push_back(EndIndex);
}
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
}
break;
}
}
}
if (Invalid || Init.isInvalid())
return ExprError();
// Clear out the expressions within the designation.
Desig.ClearExprs(*this);
DesignatedInitExpr *DIE
Use v.data() instead of &v[0] when SmallVector v might be empty. llvm-svn: 72210
2009-05-21 17:52:38 +08:00
= DesignatedInitExpr::Create(Context,
Designators.data(), Designators.size(),
InitExpressions.data(), InitExpressions.size(),
Replace more release+static_cast with takeAs. llvm-svn: 70567
2009-05-02 03:49:17 +08:00
Loc, GNUSyntax, Init.takeAs<Expr>());
Initial implementation of semantic analysis and ASTs for C99 designated initializers. This implementation should cover all of the constraints in C99 6.7.8, including long, complex designations and computing the size of incomplete array types initialized with a designated initializer. Please see the new test-case and holler if you find cases where this doesn't work. There are still some wrinkles with GNU's anonymous structs and anonymous unions (it isn't clear how these should work; we'll just follow GCC's lead) and with designated initializers for the members of a union. I'll tackle those very soon. CodeGen is still nonexistent, and there's some leftover code in the parser's representation of designators that I'll also need to clean up. llvm-svn: 62737
2009-01-22 08:58:24 +08:00
return Owned(DIE);
}
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
bool Sema::CheckInitList(const InitializedEntity &Entity,
InitListExpr *&InitList, QualType &DeclType) {
InitListChecker CheckInitList(*this, Entity, InitList, DeclType);
Move InitListChecker out of Sema.h llvm-svn: 63258
2009-01-29 08:45:39 +08:00
if (!CheckInitList.HadError())
InitList = CheckInitList.getFullyStructuredList();
return CheckInitList.HadError();
}
Check value-initializations that occur when an initializer list provides too few elements. llvm-svn: 63525
2009-02-03 01:43:21 +08:00
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
//===----------------------------------------------------------------------===//
// Initialization entity
//===----------------------------------------------------------------------===//
Improve handling of initialization by constructor, by ensuring that such initializations properly convert constructor arguments and fill in default arguments where necessary. This also makes the ownership model more clear. llvm-svn: 81394
2009-09-10 07:08:42 +08:00
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
InitializedEntity::InitializedEntity(ASTContext &Context, unsigned Index,
const InitializedEntity &Parent)
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
: Parent(&Parent), Index(Index)
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
{
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
if (const ArrayType *AT = Context.getAsArrayType(Parent.getType())) {
Kind = EK_ArrayElement;
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Type = AT->getElementType();
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
} else {
Kind = EK_VectorElement;
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Type = Parent.getType()->getAs<VectorType>()->getElementType();
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
InitializedEntity InitializedEntity::InitializeBase(ASTContext &Context,
Keep tack of whether a base in an InitializedEntity is an inherited virtual base or not. Use this in CheckConstructorAccess. llvm-svn: 102020
2010-04-22 03:52:01 +08:00
CXXBaseSpecifier *Base,
bool IsInheritedVirtualBase)
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
{
InitializedEntity Result;
Result.Kind = EK_Base;
Keep tack of whether a base in an InitializedEntity is an inherited virtual base or not. Use this in CheckConstructorAccess. llvm-svn: 102020
2010-04-22 03:52:01 +08:00
Result.Base = reinterpret_cast<uintptr_t>(Base);
if (IsInheritedVirtualBase)
Result.Base |= 0x01;
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Result.Type = Base->getType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
return Result;
}
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
DeclarationName InitializedEntity::getName() const {
switch (getKind()) {
case EK_Parameter:
Switch parameter-passing for calls via function pointers (where we don't have a FunctionDecl) over to InitializationSequence. llvm-svn: 91906
2009-12-23 00:09:06 +08:00
if (!VariableOrMember)
return DeclarationName();
// Fall through
case EK_Variable:
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
case EK_Member:
return VariableOrMember->getDeclName();
case EK_Result:
case EK_Exception:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case EK_New:
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
case EK_Temporary:
case EK_Base:
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
case EK_ArrayElement:
case EK_VectorElement:
Use MaybeCreateCXXExprWithTemporaries for potential destruction of created temporary. Use own initialized entity for copied in block variables. llvm-svn: 105533
2010-06-08 00:14:00 +08:00
case EK_BlockElement:
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
return DeclarationName();
}
// Silence GCC warning
return DeclarationName();
}
Switch more of Sema::CheckInitializerTypes over to InitializationSequence. Specially, switch initialization of a C++ class type (either copy- or direct-initialization). Also, make sure that we create an elidable copy-construction when performing copy initialization of a C++ class variable. Fixes PR5826. llvm-svn: 91750
2009-12-19 11:01:41 +08:00
DeclaratorDecl *InitializedEntity::getDecl() const {
switch (getKind()) {
case EK_Variable:
case EK_Parameter:
case EK_Member:
return VariableOrMember;
case EK_Result:
case EK_Exception:
case EK_New:
case EK_Temporary:
case EK_Base:
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
case EK_ArrayElement:
case EK_VectorElement:
Use MaybeCreateCXXExprWithTemporaries for potential destruction of created temporary. Use own initialized entity for copied in block variables. llvm-svn: 105533
2010-06-08 00:14:00 +08:00
case EK_BlockElement:
Switch more of Sema::CheckInitializerTypes over to InitializationSequence. Specially, switch initialization of a C++ class type (either copy- or direct-initialization). Also, make sure that we create an elidable copy-construction when performing copy initialization of a C++ class variable. Fixes PR5826. llvm-svn: 91750
2009-12-19 11:01:41 +08:00
return 0;
}
// Silence GCC warning
return 0;
}
Recognize when the named return value optimization applies in a "return" statement and mark the corresponding CXXConstructExpr as elidable. Teach CodeGen that eliding a temporary is different from eliding an object construction. This is just a baby step toward NRVO. llvm-svn: 103849
2010-05-15 08:13:29 +08:00
bool InitializedEntity::allowsNRVO() const {
switch (getKind()) {
case EK_Result:
case EK_Exception:
return LocAndNRVO.NRVO;
case EK_Variable:
case EK_Parameter:
case EK_Member:
case EK_New:
case EK_Temporary:
case EK_Base:
case EK_ArrayElement:
case EK_VectorElement:
Use MaybeCreateCXXExprWithTemporaries for potential destruction of created temporary. Use own initialized entity for copied in block variables. llvm-svn: 105533
2010-06-08 00:14:00 +08:00
case EK_BlockElement:
Recognize when the named return value optimization applies in a "return" statement and mark the corresponding CXXConstructExpr as elidable. Teach CodeGen that eliding a temporary is different from eliding an object construction. This is just a baby step toward NRVO. llvm-svn: 103849
2010-05-15 08:13:29 +08:00
break;
}
return false;
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
//===----------------------------------------------------------------------===//
// Initialization sequence
//===----------------------------------------------------------------------===//
void InitializationSequence::Step::Destroy() {
switch (Kind) {
case SK_ResolveAddressOfOverloadedFunction:
case SK_CastDerivedToBaseRValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_CastDerivedToBaseXValue:
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case SK_CastDerivedToBaseLValue:
case SK_BindReference:
case SK_BindReferenceToTemporary:
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
case SK_ExtraneousCopyToTemporary:
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case SK_UserConversion:
case SK_QualificationConversionRValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_QualificationConversionXValue:
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case SK_QualificationConversionLValue:
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
case SK_ListInitialization:
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
case SK_ConstructorInitialization:
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
case SK_ZeroInitialization:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case SK_CAssignment:
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
case SK_StringInit:
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
case SK_ObjCObjectConversion:
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
case SK_ConversionSequence:
delete ICS;
}
}
Switch semantic analysis of the conditional operator from using CheckReferenceInit to using the new initialization sequence code. llvm-svn: 99647
2010-03-27 04:14:36 +08:00
bool InitializationSequence::isDirectReferenceBinding() const {
return getKind() == ReferenceBinding && Steps.back().Kind == SK_BindReference;
}
bool InitializationSequence::isAmbiguous() const {
if (getKind() != FailedSequence)
return false;
switch (getFailureKind()) {
case FK_TooManyInitsForReference:
case FK_ArrayNeedsInitList:
case FK_ArrayNeedsInitListOrStringLiteral:
case FK_AddressOfOverloadFailed: // FIXME: Could do better
case FK_NonConstLValueReferenceBindingToTemporary:
case FK_NonConstLValueReferenceBindingToUnrelated:
case FK_RValueReferenceBindingToLValue:
case FK_ReferenceInitDropsQualifiers:
case FK_ReferenceInitFailed:
case FK_ConversionFailed:
case FK_TooManyInitsForScalar:
case FK_ReferenceBindingToInitList:
case FK_InitListBadDestinationType:
case FK_DefaultInitOfConst:
Add a new failure kind, FK_Incomplete, to InitializationSequence, to capture failures when we try to initialize an incomplete type. Previously, we would (ab)use FK_ConversionFailed, then occasionally dereference a null pointer when trying to diagnose the failure. Fixes <rdar://problem/7959007>. llvm-svn: 104286
2010-05-21 06:12:02 +08:00
case FK_Incomplete:
Switch semantic analysis of the conditional operator from using CheckReferenceInit to using the new initialization sequence code. llvm-svn: 99647
2010-03-27 04:14:36 +08:00
return false;
case FK_ReferenceInitOverloadFailed:
case FK_UserConversionOverloadFailed:
case FK_ConstructorOverloadFailed:
return FailedOverloadResult == OR_Ambiguous;
}
return false;
}
Collapse the three separate initialization paths in TryStaticImplicitCast (for references, class types, and everything else, respectively) into a single invocation of InitializationSequence. One of the paths (for class types) was the only client of Sema::TryInitializationByConstructor, which I have eliminated. This also simplified the interface for much of the cast-checking logic, eliminating yet more code. I've kept the representation of C++ functional casts with <> 1 arguments the same, despite the fact that I hate it. That fix will come soon. To satisfy my paranoia, I've bootstrapped + tested Clang with these changes. llvm-svn: 101549
2010-04-17 06:09:46 +08:00
bool InitializationSequence::isConstructorInitialization() const {
return !Steps.empty() && Steps.back().Kind == SK_ConstructorInitialization;
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
void InitializationSequence::AddAddressOverloadResolutionStep(
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
FunctionDecl *Function,
DeclAccessPair Found) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Step S;
S.Kind = SK_ResolveAddressOfOverloadedFunction;
S.Type = Function->getType();
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.Function.Function = Function;
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
S.Function.FoundDecl = Found;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Steps.push_back(S);
}
void InitializationSequence::AddDerivedToBaseCastStep(QualType BaseType,
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
ExprValueKind VK) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Step S;
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
switch (VK) {
case VK_RValue: S.Kind = SK_CastDerivedToBaseRValue; break;
case VK_XValue: S.Kind = SK_CastDerivedToBaseXValue; break;
case VK_LValue: S.Kind = SK_CastDerivedToBaseLValue; break;
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
default: llvm_unreachable("No such category");
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
S.Type = BaseType;
Steps.push_back(S);
}
void InitializationSequence::AddReferenceBindingStep(QualType T,
bool BindingTemporary) {
Step S;
S.Kind = BindingTemporary? SK_BindReferenceToTemporary : SK_BindReference;
S.Type = T;
Steps.push_back(S);
}
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
void InitializationSequence::AddExtraneousCopyToTemporary(QualType T) {
Step S;
S.Kind = SK_ExtraneousCopyToTemporary;
S.Type = T;
Steps.push_back(S);
}
Fix a recent regression from the initialization changes. llvm-svn: 91097
2009-12-11 10:42:07 +08:00
void InitializationSequence::AddUserConversionStep(FunctionDecl *Function,
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
DeclAccessPair FoundDecl,
Fix a recent regression from the initialization changes. llvm-svn: 91097
2009-12-11 10:42:07 +08:00
QualType T) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Step S;
S.Kind = SK_UserConversion;
Fix a recent regression from the initialization changes. llvm-svn: 91097
2009-12-11 10:42:07 +08:00
S.Type = T;
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.Function.Function = Function;
S.Function.FoundDecl = FoundDecl;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Steps.push_back(S);
}
void InitializationSequence::AddQualificationConversionStep(QualType Ty,
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
ExprValueKind VK) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Step S;
Work around a gcc warning. llvm-svn: 112192
2010-08-27 00:36:35 +08:00
S.Kind = SK_QualificationConversionRValue; // work around a gcc warning
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
switch (VK) {
case VK_RValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
S.Kind = SK_QualificationConversionRValue;
break;
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
case VK_XValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
S.Kind = SK_QualificationConversionXValue;
break;
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
case VK_LValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
S.Kind = SK_QualificationConversionLValue;
break;
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
S.Type = Ty;
Steps.push_back(S);
}
void InitializationSequence::AddConversionSequenceStep(
const ImplicitConversionSequence &ICS,
QualType T) {
Step S;
S.Kind = SK_ConversionSequence;
S.Type = T;
S.ICS = new ImplicitConversionSequence(ICS);
Steps.push_back(S);
}
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
void InitializationSequence::AddListInitializationStep(QualType T) {
Step S;
S.Kind = SK_ListInitialization;
S.Type = T;
Steps.push_back(S);
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
void
InitializationSequence::AddConstructorInitializationStep(
CXXConstructorDecl *Constructor,
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
AccessSpecifier Access,
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
QualType T) {
Step S;
S.Kind = SK_ConstructorInitialization;
S.Type = T;
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.Function.Function = Constructor;
S.Function.FoundDecl = DeclAccessPair::make(Constructor, Access);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
Steps.push_back(S);
}
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
void InitializationSequence::AddZeroInitializationStep(QualType T) {
Step S;
S.Kind = SK_ZeroInitialization;
S.Type = T;
Steps.push_back(S);
}
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
void InitializationSequence::AddCAssignmentStep(QualType T) {
Step S;
S.Kind = SK_CAssignment;
S.Type = T;
Steps.push_back(S);
}
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
void InitializationSequence::AddStringInitStep(QualType T) {
Step S;
S.Kind = SK_StringInit;
S.Type = T;
Steps.push_back(S);
}
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
void InitializationSequence::AddObjCObjectConversionStep(QualType T) {
Step S;
S.Kind = SK_ObjCObjectConversion;
S.Type = T;
Steps.push_back(S);
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
void InitializationSequence::SetOverloadFailure(FailureKind Failure,
OverloadingResult Result) {
SequenceKind = FailedSequence;
this->Failure = Failure;
this->FailedOverloadResult = Result;
}
//===----------------------------------------------------------------------===//
// Attempt initialization
//===----------------------------------------------------------------------===//
/// \brief Attempt list initialization (C++0x [dcl.init.list])
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
static void TryListInitialization(Sema &S,
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
const InitializedEntity &Entity,
const InitializationKind &Kind,
InitListExpr *InitList,
InitializationSequence &Sequence) {
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
// FIXME: We only perform rudimentary checking of list
// initializations at this point, then assume that any list
// initialization of an array, aggregate, or scalar will be
Fix a comment typo. llvm-svn: 107290
2010-07-01 00:41:54 +08:00
// well-formed. When we actually "perform" list initialization, we'll
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
// do all of the necessary checking. C++0x initializer lists will
// force us to perform more checking here.
Sequence.setSequenceKind(InitializationSequence::ListInitialization);
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType();
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
// C++ [dcl.init]p13:
// If T is a scalar type, then a declaration of the form
//
// T x = { a };
//
// is equivalent to
//
// T x = a;
if (DestType->isScalarType()) {
if (InitList->getNumInits() > 1 && S.getLangOptions().CPlusPlus) {
Sequence.SetFailed(InitializationSequence::FK_TooManyInitsForScalar);
return;
}
// Assume scalar initialization from a single value works.
} else if (DestType->isAggregateType()) {
// Assume aggregate initialization works.
} else if (DestType->isVectorType()) {
// Assume vector initialization works.
} else if (DestType->isReferenceType()) {
// FIXME: C++0x defines behavior for this.
Sequence.SetFailed(InitializationSequence::FK_ReferenceBindingToInitList);
return;
} else if (DestType->isRecordType()) {
// FIXME: C++0x defines behavior for this
Sequence.SetFailed(InitializationSequence::FK_InitListBadDestinationType);
}
// Add a general "list initialization" step.
Sequence.AddListInitializationStep(DestType);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
/// \brief Try a reference initialization that involves calling a conversion
/// function.
static OverloadingResult TryRefInitWithConversionFunction(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
Expr *Initializer,
bool AllowRValues,
InitializationSequence &Sequence) {
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
QualType cv1T1 = DestType->getAs<ReferenceType>()->getPointeeType();
QualType T1 = cv1T1.getUnqualifiedType();
QualType cv2T2 = Initializer->getType();
QualType T2 = cv2T2.getUnqualifiedType();
bool DerivedToBase;
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
bool ObjCConversion;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
assert(!S.CompareReferenceRelationship(Initializer->getLocStart(),
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
T1, T2, DerivedToBase,
ObjCConversion) &&
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
"Must have incompatible references when binding via conversion");
Silence some release build warnings. llvm-svn: 91234
2009-12-13 09:37:04 +08:00
(void)DerivedToBase;
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
(void)ObjCConversion;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Build the candidate set directly in the initialization sequence
// structure, so that it will persist if we fail.
OverloadCandidateSet &CandidateSet = Sequence.getFailedCandidateSet();
CandidateSet.clear();
// Determine whether we are allowed to call explicit constructors or
// explicit conversion operators.
bool AllowExplicit = Kind.getKind() == InitializationKind::IK_Direct;
const RecordType *T1RecordType = 0;
Reapply the reference-binding patch applied below, along with a fix to ensure that we complete the type when we need to look at constructors during reference binding. When determining whether the two types involved in reference binding are reference-compatible, reference-related, etc., do not complete the type of the reference itself because it is not necessary to determine well-formedness of the program. Complete the type that we are binding to, since that can affect whether we know about a derived-to-base conversion. Re-fixes PR7080. llvm-svn: 103283
2010-05-08 03:42:26 +08:00
if (AllowRValues && (T1RecordType = T1->getAs<RecordType>()) &&
!S.RequireCompleteType(Kind.getLocation(), T1, 0)) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// The type we're converting to is a class type. Enumerate its constructors
// to see if there is a suitable conversion.
CXXRecordDecl *T1RecordDecl = cast<CXXRecordDecl>(T1RecordType->getDecl());
Don't try to initialize a reference with a constructed temporary if either of the classes is invalid. A class is invalid if a base is invalid. llvm-svn: 111227
2010-08-17 15:23:57 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
DeclContext::lookup_iterator Con, ConEnd;
Introduce a new routine, LookupConstructors(), and use it for all constructor-name lookup. llvm-svn: 107536
2010-07-03 07:12:18 +08:00
for (llvm::tie(Con, ConEnd) = S.LookupConstructors(T1RecordDecl);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Con != ConEnd; ++Con) {
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
NamedDecl *D = *Con;
DeclAccessPair FoundDecl = DeclAccessPair::make(D, D->getAccess());
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Find the constructor (which may be a template).
CXXConstructorDecl *Constructor = 0;
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
FunctionTemplateDecl *ConstructorTmpl = dyn_cast<FunctionTemplateDecl>(D);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (ConstructorTmpl)
Constructor = cast<CXXConstructorDecl>(
ConstructorTmpl->getTemplatedDecl());
else
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
Constructor = cast<CXXConstructorDecl>(D);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (!Constructor->isInvalidDecl() &&
Constructor->isConvertingConstructor(AllowExplicit)) {
if (ConstructorTmpl)
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddTemplateOverloadCandidate(ConstructorTmpl, FoundDecl,
Pass access specifiers around in overload resolution. llvm-svn: 94485
2010-01-26 09:37:31 +08:00
/*ExplicitArgs*/ 0,
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
&Initializer, 1, CandidateSet);
else
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddOverloadCandidate(Constructor, FoundDecl,
Pass access specifiers around in overload resolution. llvm-svn: 94485
2010-01-26 09:37:31 +08:00
&Initializer, 1, CandidateSet);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
}
}
Don't try to initialize a reference with a constructed temporary if either of the classes is invalid. A class is invalid if a base is invalid. llvm-svn: 111227
2010-08-17 15:23:57 +08:00
if (T1RecordType && T1RecordType->getDecl()->isInvalidDecl())
return OR_No_Viable_Function;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Reapply the reference-binding patch applied below, along with a fix to ensure that we complete the type when we need to look at constructors during reference binding. When determining whether the two types involved in reference binding are reference-compatible, reference-related, etc., do not complete the type of the reference itself because it is not necessary to determine well-formedness of the program. Complete the type that we are binding to, since that can affect whether we know about a derived-to-base conversion. Re-fixes PR7080. llvm-svn: 103283
2010-05-08 03:42:26 +08:00
const RecordType *T2RecordType = 0;
if ((T2RecordType = T2->getAs<RecordType>()) &&
!S.RequireCompleteType(Kind.getLocation(), T2, 0)) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// The type we're converting from is a class type, enumerate its conversion
// functions.
CXXRecordDecl *T2RecordDecl = cast<CXXRecordDecl>(T2RecordType->getDecl());
// Determine the type we are converting to. If we are allowed to
// convert to an rvalue, take the type that the destination type
// refers to.
QualType ToType = AllowRValues? cv1T1 : DestType;
Give UnresolvedSet the ability to store access specifiers for each declaration. Change LookupResult to use UnresolvedSet. Also extract UnresolvedSet into its own header and make it templated over an inline capacity. llvm-svn: 93959
2010-01-20 08:46:10 +08:00
const UnresolvedSetImpl *Conversions
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
= T2RecordDecl->getVisibleConversionFunctions();
Give UnresolvedSet the ability to store access specifiers for each declaration. Change LookupResult to use UnresolvedSet. Also extract UnresolvedSet into its own header and make it templated over an inline capacity. llvm-svn: 93959
2010-01-20 08:46:10 +08:00
for (UnresolvedSetImpl::const_iterator I = Conversions->begin(),
E = Conversions->end(); I != E; ++I) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
NamedDecl *D = *I;
CXXRecordDecl *ActingDC = cast<CXXRecordDecl>(D->getDeclContext());
if (isa<UsingShadowDecl>(D))
D = cast<UsingShadowDecl>(D)->getTargetDecl();
FunctionTemplateDecl *ConvTemplate = dyn_cast<FunctionTemplateDecl>(D);
CXXConversionDecl *Conv;
if (ConvTemplate)
Conv = cast<CXXConversionDecl>(ConvTemplate->getTemplatedDecl());
else
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Conv = cast<CXXConversionDecl>(D);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// If the conversion function doesn't return a reference type,
// it can't be considered for this conversion unless we're allowed to
// consider rvalues.
// FIXME: Do we need to make sure that we only consider conversion
// candidates with reference-compatible results? That might be needed to
// break recursion.
if ((AllowExplicit || !Conv->isExplicit()) &&
(AllowRValues || Conv->getConversionType()->isLValueReferenceType())){
if (ConvTemplate)
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddTemplateConversionCandidate(ConvTemplate, I.getPair(),
Pass access specifiers around in overload resolution. llvm-svn: 94485
2010-01-26 09:37:31 +08:00
ActingDC, Initializer,
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
ToType, CandidateSet);
else
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddConversionCandidate(Conv, I.getPair(), ActingDC,
Commit Eli's fix for implicit conversions to array type. Fixes PR6264. llvm-svn: 97202
2010-02-26 09:17:27 +08:00
Initializer, ToType, CandidateSet);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
}
}
Don't try to initialize a reference with a constructed temporary if either of the classes is invalid. A class is invalid if a base is invalid. llvm-svn: 111227
2010-08-17 15:23:57 +08:00
if (T2RecordType && T2RecordType->getDecl()->isInvalidDecl())
return OR_No_Viable_Function;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
SourceLocation DeclLoc = Initializer->getLocStart();
// Perform overload resolution. If it fails, return the failed result.
OverloadCandidateSet::iterator Best;
if (OverloadingResult Result
When performing overload resolution, only compare the final conversion sequences for two conversion functions when in fact we are in the text of initialization by a user-defined conversion sequences. Fixes PR8034. llvm-svn: 113724
2010-09-12 16:07:23 +08:00
= CandidateSet.BestViableFunction(S, DeclLoc, Best, true))
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
return Result;
Fix a recent regression from the initialization changes. llvm-svn: 91097
2009-12-11 10:42:07 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
FunctionDecl *Function = Best->Function;
Fix a recent regression from the initialization changes. llvm-svn: 91097
2009-12-11 10:42:07 +08:00
// Compute the returned type of the conversion.
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (isa<CXXConversionDecl>(Function))
T2 = Function->getResultType();
else
T2 = cv1T1;
Fix a recent regression from the initialization changes. llvm-svn: 91097
2009-12-11 10:42:07 +08:00
// Add the user-defined conversion step.
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
Sequence.AddUserConversionStep(Function, Best->FoundDecl,
Whenever we're creating an expression that is typically an rvalue (e.g., a call, cast, etc.), immediately adjust the expression's type to strip cv-qualifiers off of all non-class types (in C++) or all types (in C). This effectively extends my previous fix for PR7463, which was restricted to calls, to other kinds of expressions within similar characteristics. I've audited every use of getNonReferenceType() in the code base, switching to the newly-renamed getNonLValueExprType() where necessary. Big thanks to Eli for pointing out just how incomplete my original fix for PR7463 actually was. We've been handling cv-qualifiers on rvalues wrong for a very, very long time. Fixes PR7463. llvm-svn: 108253
2010-07-14 02:40:04 +08:00
T2.getNonLValueExprType(S.Context));
Fix a recent regression from the initialization changes. llvm-svn: 91097
2009-12-11 10:42:07 +08:00
// Determine whether we need to perform derived-to-base or
// cv-qualification adjustments.
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
ExprValueKind VK = VK_RValue;
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
if (T2->isLValueReferenceType())
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK = VK_LValue;
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
else if (const RValueReferenceType *RRef = T2->getAs<RValueReferenceType>())
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK = RRef->getPointeeType()->isFunctionType() ? VK_LValue : VK_XValue;
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
bool NewDerivedToBase = false;
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
bool NewObjCConversion = false;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Sema::ReferenceCompareResult NewRefRelationship
Whenever we're creating an expression that is typically an rvalue (e.g., a call, cast, etc.), immediately adjust the expression's type to strip cv-qualifiers off of all non-class types (in C++) or all types (in C). This effectively extends my previous fix for PR7463, which was restricted to calls, to other kinds of expressions within similar characteristics. I've audited every use of getNonReferenceType() in the code base, switching to the newly-renamed getNonLValueExprType() where necessary. Big thanks to Eli for pointing out just how incomplete my original fix for PR7463 actually was. We've been handling cv-qualifiers on rvalues wrong for a very, very long time. Fixes PR7463. llvm-svn: 108253
2010-07-14 02:40:04 +08:00
= S.CompareReferenceRelationship(DeclLoc, T1,
T2.getNonLValueExprType(S.Context),
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
NewDerivedToBase, NewObjCConversion);
Reference binding via user-defined conversion can compute a binding that is not reference-related (because it requires another implicit conversion to which we can find). Fixes PR6483. llvm-svn: 97922
2010-03-08 07:17:44 +08:00
if (NewRefRelationship == Sema::Ref_Incompatible) {
// If the type we've converted to is not reference-related to the
// type we're looking for, then there is another conversion step
// we need to perform to produce a temporary of the right type
// that we'll be binding to.
ImplicitConversionSequence ICS;
ICS.setStandard();
ICS.Standard = Best->FinalConversion;
T2 = ICS.Standard.getToType(2);
Sequence.AddConversionSequenceStep(ICS, T2);
} else if (NewDerivedToBase)
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Sequence.AddDerivedToBaseCastStep(
S.Context.getQualifiedType(T1,
T2.getNonReferenceType().getQualifiers()),
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK);
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
else if (NewObjCConversion)
Sequence.AddObjCObjectConversionStep(
S.Context.getQualifiedType(T1,
T2.getNonReferenceType().getQualifiers()));
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (cv1T1.getQualifiers() != T2.getNonReferenceType().getQualifiers())
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
Sequence.AddQualificationConversionStep(cv1T1, VK);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Sequence.AddReferenceBindingStep(cv1T1, !T2->isReferenceType());
return OR_Success;
}
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
/// \brief Attempt reference initialization (C++0x [dcl.init.ref])
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
static void TryReferenceInitialization(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
Expr *Initializer,
InitializationSequence &Sequence) {
Sequence.setSequenceKind(InitializationSequence::ReferenceBinding);
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
QualType cv1T1 = DestType->getAs<ReferenceType>()->getPointeeType();
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
Qualifiers T1Quals;
QualType T1 = S.Context.getUnqualifiedArrayType(cv1T1, T1Quals);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
QualType cv2T2 = Initializer->getType();
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
Qualifiers T2Quals;
QualType T2 = S.Context.getUnqualifiedArrayType(cv2T2, T2Quals);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
SourceLocation DeclLoc = Initializer->getLocStart();
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// If the initializer is the address of an overloaded function, try
// to resolve the overloaded function. If all goes well, T2 is the
// type of the resulting function.
if (S.Context.getCanonicalType(T2) == S.Context.OverloadTy) {
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
DeclAccessPair Found;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
FunctionDecl *Fn = S.ResolveAddressOfOverloadedFunction(Initializer,
T1,
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
false,
Found);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (!Fn) {
Sequence.SetFailed(InitializationSequence::FK_AddressOfOverloadFailed);
return;
}
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
Sequence.AddAddressOverloadResolutionStep(Fn, Found);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
cv2T2 = Fn->getType();
T2 = cv2T2.getUnqualifiedType();
}
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Compute some basic properties of the types and the initializer.
bool isLValueRef = DestType->isLValueReferenceType();
bool isRValueRef = !isLValueRef;
bool DerivedToBase = false;
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
bool ObjCConversion = false;
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Expr::Classification InitCategory = Initializer->Classify(S.Context);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Sema::ReferenceCompareResult RefRelationship
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
= S.CompareReferenceRelationship(DeclLoc, cv1T1, cv2T2, DerivedToBase,
ObjCConversion);
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// C++0x [dcl.init.ref]p5:
// A reference to type "cv1 T1" is initialized by an expression of type
// "cv2 T2" as follows:
//
// - If the reference is an lvalue reference and the initializer
// expression
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
// Note the analogous bullet points for rvlaue refs to functions. Because
// there are no function rvalues in C++, rvalue refs to functions are treated
// like lvalue refs.
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
OverloadingResult ConvOvlResult = OR_Success;
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
bool T1Function = T1->isFunctionType();
if (isLValueRef || T1Function) {
if (InitCategory.isLValue() &&
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
RefRelationship >= Sema::Ref_Compatible_With_Added_Qualification) {
// - is an lvalue (but is not a bit-field), and "cv1 T1" is
// reference-compatible with "cv2 T2," or
//
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
// Per C++ [over.best.ics]p2, we don't diagnose whether the lvalue is a
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// bit-field when we're determining whether the reference initialization
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
// can occur. However, we do pay attention to whether it is a bit-field
// to decide whether we're actually binding to a temporary created from
// the bit-field.
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (DerivedToBase)
Sequence.AddDerivedToBaseCastStep(
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
S.Context.getQualifiedType(T1, T2Quals),
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK_LValue);
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
else if (ObjCConversion)
Sequence.AddObjCObjectConversionStep(
S.Context.getQualifiedType(T1, T2Quals));
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
if (T1Quals != T2Quals)
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
Sequence.AddQualificationConversionStep(cv1T1, VK_LValue);
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
bool BindingTemporary = T1Quals.hasConst() && !T1Quals.hasVolatile() &&
Diagnose binding a non-const reference to a vector element. llvm-svn: 94963
2010-02-01 01:18:49 +08:00
(Initializer->getBitField() || Initializer->refersToVectorElement());
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
Sequence.AddReferenceBindingStep(cv1T1, BindingTemporary);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
return;
}
// - has a class type (i.e., T2 is a class type), where T1 is not
// reference-related to T2, and can be implicitly converted to an
// lvalue of type "cv3 T3," where "cv1 T1" is reference-compatible
// with "cv3 T3" (this conversion is selected by enumerating the
// applicable conversion functions (13.3.1.6) and choosing the best
// one through overload resolution (13.3)),
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
// If we have an rvalue ref to function type here, the rhs must be
// an rvalue.
if (RefRelationship == Sema::Ref_Incompatible && T2->isRecordType() &&
(isLValueRef || InitCategory.isRValue())) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
ConvOvlResult = TryRefInitWithConversionFunction(S, Entity, Kind,
Initializer,
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
/*AllowRValues=*/isRValueRef,
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Sequence);
if (ConvOvlResult == OR_Success)
return;
Introduce a specific representation for the ambiguous implicit conversion sequence. Lots of small relevant changes. Fixes some serious problems with ambiguous conversions; also possibly improves associated diagnostics. llvm-svn: 93214
2010-01-12 08:44:57 +08:00
if (ConvOvlResult != OR_No_Viable_Function) {
Sequence.SetOverloadFailure(
InitializationSequence::FK_ReferenceInitOverloadFailed,
ConvOvlResult);
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
}
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// - Otherwise, the reference shall be an lvalue reference to a
// non-volatile const type (i.e., cv1 shall be const), or the reference
// shall be an rvalue reference and the initializer expression shall
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
// be an rvalue or have a function type.
// We handled the function type stuff above.
Fix reference-binding when we have a reference to const volatile type; previously, we were allowing this to bind to a temporary. Now, we don't; add test-cases and improve diagnostics. llvm-svn: 94831
2010-01-30 03:39:15 +08:00
if (!((isLValueRef && T1Quals.hasConst() && !T1Quals.hasVolatile()) ||
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
(isRValueRef && InitCategory.isRValue()))) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (ConvOvlResult && !Sequence.getFailedCandidateSet().empty())
Sequence.SetOverloadFailure(
InitializationSequence::FK_ReferenceInitOverloadFailed,
ConvOvlResult);
else if (isLValueRef)
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Sequence.SetFailed(InitCategory.isLValue()
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
? (RefRelationship == Sema::Ref_Related
? InitializationSequence::FK_ReferenceInitDropsQualifiers
: InitializationSequence::FK_NonConstLValueReferenceBindingToUnrelated)
: InitializationSequence::FK_NonConstLValueReferenceBindingToTemporary);
else
Sequence.SetFailed(
InitializationSequence::FK_RValueReferenceBindingToLValue);
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
return;
}
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
// - [If T1 is not a function type], if T2 is a class type and
if (!T1Function && T2->isRecordType()) {
Make sure that implicit qualification and derived-to-base conversions of xvalues preserve xvalue-ness. Unfortunately I have no idea how to test this property; there doesn't seem to be a syntactical construct that triggers such a conversion and still allows the distinction between prvalues and xvalues to be made. llvm-svn: 109406
2010-07-27 01:52:21 +08:00
bool isXValue = InitCategory.isXValue();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// - the initializer expression is an rvalue and "cv1 T1" is
// reference-compatible with "cv2 T2", or
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
if (InitCategory.isRValue() &&
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
RefRelationship >= Sema::Ref_Compatible_With_Added_Qualification) {
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
// The corresponding bullet in C++03 [dcl.init.ref]p5 gives the
// compiler the freedom to perform a copy here or bind to the
// object, while C++0x requires that we bind directly to the
// object. Hence, we always bind to the object without making an
// extra copy. However, in C++03 requires that we check for the
// presence of a suitable copy constructor:
//
// The constructor that would be used to make the copy shall
// be callable whether or not the copy is actually done.
if (!S.getLangOptions().CPlusPlus0x)
Sequence.AddExtraneousCopyToTemporary(cv2T2);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (DerivedToBase)
Sequence.AddDerivedToBaseCastStep(
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
S.Context.getQualifiedType(T1, T2Quals),
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
isXValue ? VK_XValue : VK_RValue);
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
else if (ObjCConversion)
Sequence.AddObjCObjectConversionStep(
S.Context.getQualifiedType(T1, T2Quals));
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
if (T1Quals != T2Quals)
Make sure that implicit qualification and derived-to-base conversions of xvalues preserve xvalue-ness. Unfortunately I have no idea how to test this property; there doesn't seem to be a syntactical construct that triggers such a conversion and still allows the distinction between prvalues and xvalues to be made. llvm-svn: 109406
2010-07-27 01:52:21 +08:00
Sequence.AddQualificationConversionStep(cv1T1,
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
isXValue ? VK_XValue : VK_RValue);
Make sure that implicit qualification and derived-to-base conversions of xvalues preserve xvalue-ness. Unfortunately I have no idea how to test this property; there doesn't seem to be a syntactical construct that triggers such a conversion and still allows the distinction between prvalues and xvalues to be made. llvm-svn: 109406
2010-07-27 01:52:21 +08:00
Sequence.AddReferenceBindingStep(cv1T1, /*bindingTemporary=*/!isXValue);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
return;
}
Make both old and new versions of reference binding use the new classification functions, and updated them for N3092. llvm-svn: 107301
2010-07-01 02:13:39 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// - T1 is not reference-related to T2 and the initializer expression
// can be implicitly converted to an rvalue of type "cv3 T3" (this
// conversion is selected by enumerating the applicable conversion
// functions (13.3.1.6) and choosing the best one through overload
// resolution (13.3)),
if (RefRelationship == Sema::Ref_Incompatible) {
ConvOvlResult = TryRefInitWithConversionFunction(S, Entity,
Kind, Initializer,
/*AllowRValues=*/true,
Sequence);
if (ConvOvlResult)
Sequence.SetOverloadFailure(
InitializationSequence::FK_ReferenceInitOverloadFailed,
ConvOvlResult);
return;
}
Sequence.SetFailed(InitializationSequence::FK_ReferenceInitDropsQualifiers);
return;
}
// - If the initializer expression is an rvalue, with T2 an array type,
// and "cv1 T1" is reference-compatible with "cv2 T2," the reference
// is bound to the object represented by the rvalue (see 3.10).
// FIXME: How can an array type be reference-compatible with anything?
// Don't we mean the element types of T1 and T2?
// - Otherwise, a temporary of type “cv1 T1” is created and initialized
// from the initializer expression using the rules for a non-reference
// copy initialization (8.5). The reference is then bound to the
// temporary. [...]
More refactoring. llvm-svn: 105458
2010-06-04 10:29:22 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Determine whether we are allowed to call explicit constructors or
// explicit conversion operators.
bool AllowExplicit = (Kind.getKind() == InitializationKind::IK_Direct);
More refactoring. llvm-svn: 105458
2010-06-04 10:29:22 +08:00
InitializedEntity TempEntity = InitializedEntity::InitializeTemporary(cv1T1);
if (S.TryImplicitConversion(Sequence, TempEntity, Initializer,
/*SuppressUserConversions*/ false,
AllowExplicit,
/*FIXME:InOverloadResolution=*/false)) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// FIXME: Use the conversion function set stored in ICS to turn
// this into an overloading ambiguity diagnostic. However, we need
// to keep that set as an OverloadCandidateSet rather than as some
// other kind of set.
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
if (ConvOvlResult && !Sequence.getFailedCandidateSet().empty())
Sequence.SetOverloadFailure(
InitializationSequence::FK_ReferenceInitOverloadFailed,
ConvOvlResult);
else
Sequence.SetFailed(InitializationSequence::FK_ReferenceInitFailed);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
return;
}
// [...] If T1 is reference-related to T2, cv1 must be the
// same cv-qualification as, or greater cv-qualification
// than, cv2; otherwise, the program is ill-formed.
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
unsigned T1CVRQuals = T1Quals.getCVRQualifiers();
unsigned T2CVRQuals = T2Quals.getCVRQualifiers();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (RefRelationship == Sema::Ref_Related &&
Fix the CodeGen half of PR5911 by changing reference initialization to correctly look through arrays to see cv-qualifiers. Also enhances the routine for doing this to preserve more type sugaring for diagnostics. llvm-svn: 93252
2010-01-13 04:32:25 +08:00
(T1CVRQuals | T2CVRQuals) != T1CVRQuals) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Sequence.SetFailed(InitializationSequence::FK_ReferenceInitDropsQualifiers);
return;
}
Sequence.AddReferenceBindingStep(cv1T1, /*bindingTemporary=*/true);
return;
}
/// \brief Attempt character array initialization from a string literal
/// (C++ [dcl.init.string], C99 6.7.8).
static void TryStringLiteralInitialization(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
Expr *Initializer,
InitializationSequence &Sequence) {
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
Sequence.setSequenceKind(InitializationSequence::StringInit);
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Sequence.AddStringInitStep(Entity.getType());
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
/// \brief Attempt initialization by constructor (C++ [dcl.init]), which
/// enumerates the constructors of the initialized entity and performs overload
/// resolution to select the best.
static void TryConstructorInitialization(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
Expr **Args, unsigned NumArgs,
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
QualType DestType,
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
InitializationSequence &Sequence) {
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
Sequence.setSequenceKind(InitializationSequence::ConstructorInitialization);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
// Build the candidate set directly in the initialization sequence
// structure, so that it will persist if we fail.
OverloadCandidateSet &CandidateSet = Sequence.getFailedCandidateSet();
CandidateSet.clear();
// Determine whether we are allowed to call explicit constructors or
// explicit conversion operators.
bool AllowExplicit = (Kind.getKind() == InitializationKind::IK_Direct ||
Kind.getKind() == InitializationKind::IK_Value ||
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
Kind.getKind() == InitializationKind::IK_Default);
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
// The type we're constructing needs to be complete.
if (S.RequireCompleteType(Kind.getLocation(), DestType, 0)) {
Add a new failure kind, FK_Incomplete, to InitializationSequence, to capture failures when we try to initialize an incomplete type. Previously, we would (ab)use FK_ConversionFailed, then occasionally dereference a null pointer when trying to diagnose the failure. Fixes <rdar://problem/7959007>. llvm-svn: 104286
2010-05-21 06:12:02 +08:00
Sequence.SetFailed(InitializationSequence::FK_Incomplete);
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
return;
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
// The type we're converting to is a class type. Enumerate its constructors
// to see if one is suitable.
const RecordType *DestRecordType = DestType->getAs<RecordType>();
assert(DestRecordType && "Constructor initialization requires record type");
CXXRecordDecl *DestRecordDecl
= cast<CXXRecordDecl>(DestRecordType->getDecl());
DeclContext::lookup_iterator Con, ConEnd;
Introduce a new routine, LookupConstructors(), and use it for all constructor-name lookup. llvm-svn: 107536
2010-07-03 07:12:18 +08:00
for (llvm::tie(Con, ConEnd) = S.LookupConstructors(DestRecordDecl);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
Con != ConEnd; ++Con) {
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
NamedDecl *D = *Con;
DeclAccessPair FoundDecl = DeclAccessPair::make(D, D->getAccess());
When we are performing copy initialization of a class type via its copy constructor, suppress user-defined conversions on the argument. Otherwise, we can end up in a recursion loop where the bind the argument of the copy constructor to another copy constructor call, whose argument is then a copy constructor call... Found by Boost.Regex which, alas, still isn't building. llvm-svn: 102269
2010-04-25 04:54:38 +08:00
bool SuppressUserConversions = false;
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
// Find the constructor (which may be a template).
CXXConstructorDecl *Constructor = 0;
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
FunctionTemplateDecl *ConstructorTmpl = dyn_cast<FunctionTemplateDecl>(D);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
if (ConstructorTmpl)
Constructor = cast<CXXConstructorDecl>(
ConstructorTmpl->getTemplatedDecl());
When we are performing copy initialization of a class type via its copy constructor, suppress user-defined conversions on the argument. Otherwise, we can end up in a recursion loop where the bind the argument of the copy constructor to another copy constructor call, whose argument is then a copy constructor call... Found by Boost.Regex which, alas, still isn't building. llvm-svn: 102269
2010-04-25 04:54:38 +08:00
else {
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
Constructor = cast<CXXConstructorDecl>(D);
When we are performing copy initialization of a class type via its copy constructor, suppress user-defined conversions on the argument. Otherwise, we can end up in a recursion loop where the bind the argument of the copy constructor to another copy constructor call, whose argument is then a copy constructor call... Found by Boost.Regex which, alas, still isn't building. llvm-svn: 102269
2010-04-25 04:54:38 +08:00
// If we're performing copy initialization using a copy constructor, we
// suppress user-defined conversions on the arguments.
// FIXME: Move constructors?
if (Kind.getKind() == InitializationKind::IK_Copy &&
Constructor->isCopyConstructor())
SuppressUserConversions = true;
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
if (!Constructor->isInvalidDecl() &&
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
(AllowExplicit || !Constructor->isExplicit())) {
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
if (ConstructorTmpl)
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddTemplateOverloadCandidate(ConstructorTmpl, FoundDecl,
Pass access specifiers around in overload resolution. llvm-svn: 94485
2010-01-26 09:37:31 +08:00
/*ExplicitArgs*/ 0,
When we are performing copy initialization of a class type via its copy constructor, suppress user-defined conversions on the argument. Otherwise, we can end up in a recursion loop where the bind the argument of the copy constructor to another copy constructor call, whose argument is then a copy constructor call... Found by Boost.Regex which, alas, still isn't building. llvm-svn: 102269
2010-04-25 04:54:38 +08:00
Args, NumArgs, CandidateSet,
SuppressUserConversions);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
else
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddOverloadCandidate(Constructor, FoundDecl,
When we are performing copy initialization of a class type via its copy constructor, suppress user-defined conversions on the argument. Otherwise, we can end up in a recursion loop where the bind the argument of the copy constructor to another copy constructor call, whose argument is then a copy constructor call... Found by Boost.Regex which, alas, still isn't building. llvm-svn: 102269
2010-04-25 04:54:38 +08:00
Args, NumArgs, CandidateSet,
SuppressUserConversions);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
}
}
SourceLocation DeclLoc = Kind.getLocation();
// Perform overload resolution. If it fails, return the failed result.
OverloadCandidateSet::iterator Best;
if (OverloadingResult Result
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
= CandidateSet.BestViableFunction(S, DeclLoc, Best)) {
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
Sequence.SetOverloadFailure(
InitializationSequence::FK_ConstructorOverloadFailed,
Result);
return;
}
Rework base and member initialization in constructors, with several (necessarily simultaneous) changes: - CXXBaseOrMemberInitializer now contains only a single initializer rather than a set of initialiation arguments + a constructor. The single initializer covers all aspects of initialization, including constructor calls as necessary but also cleanup of temporaries created by the initializer (which we never handled before!). - Rework + simplify code generation for CXXBaseOrMemberInitializers, since we can now just emit the initializer as an initializer. - Switched base and member initialization over to the new initialization code (InitializationSequence), so that it - Improved diagnostics for the new initialization code when initializing bases and members, to match the diagnostics produced by the previous (special-purpose) code. - Simplify the representation of type-checked constructor initializers in templates; instead of keeping the fully-type-checked AST, which is rather hard to undo at template instantiation time, throw away the type-checked AST and store the raw expressions in the AST. This simplifies instantiation, but loses a little but of information in the AST. - When type-checking implicit base or member initializers within a dependent context, don't add the generated initializers into the AST, because they'll look like they were explicit. - Record in CXXConstructExpr when the constructor call is to initialize a base class, so that CodeGen does not have to infer it from context. This ensures that we call the right kind of constructor. There are also a few "opportunity" fixes here that were needed to not regress, for example: - Diagnose default-initialization of a const-qualified class that does not have a user-declared default constructor. We had this diagnostic specifically for bases and members, but missed it for variables. That's fixed now. - When defining the implicit constructors, destructor, and copy-assignment operator, set the CurContext to that constructor when we're defining the body. llvm-svn: 94952
2010-01-31 17:12:51 +08:00
// C++0x [dcl.init]p6:
// If a program calls for the default initialization of an object
// of a const-qualified type T, T shall be a class type with a
// user-provided default constructor.
if (Kind.getKind() == InitializationKind::IK_Default &&
Entity.getType().isConstQualified() &&
cast<CXXConstructorDecl>(Best->Function)->isImplicit()) {
Sequence.SetFailed(InitializationSequence::FK_DefaultInitOfConst);
return;
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
// Add the constructor initialization step. Any cv-qualification conversion is
// subsumed by the initialization.
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
Sequence.AddConstructorInitializationStep(
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
cast<CXXConstructorDecl>(Best->Function),
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
Best->FoundDecl.getAccess(),
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
DestType);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
/// \brief Attempt value initialization (C++ [dcl.init]p7).
static void TryValueInitialization(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
InitializationSequence &Sequence) {
// C++ [dcl.init]p5:
//
// To value-initialize an object of type T means:
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType T = Entity.getType();
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
// -- if T is an array type, then each element is value-initialized;
while (const ArrayType *AT = S.Context.getAsArrayType(T))
T = AT->getElementType();
if (const RecordType *RT = T->getAs<RecordType>()) {
if (CXXRecordDecl *ClassDecl = dyn_cast<CXXRecordDecl>(RT->getDecl())) {
// -- if T is a class type (clause 9) with a user-declared
// constructor (12.1), then the default constructor for T is
// called (and the initialization is ill-formed if T has no
// accessible default constructor);
//
// FIXME: we really want to refer to a single subobject of the array,
// but Entity doesn't have a way to capture that (yet).
if (ClassDecl->hasUserDeclaredConstructor())
return TryConstructorInitialization(S, Entity, Kind, 0, 0, T, Sequence);
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
// -- if T is a (possibly cv-qualified) non-union class type
// without a user-provided constructor, then the object is
// zero-initialized and, if T’s implicitly-declared default
// constructor is non-trivial, that constructor is called.
Merged Elaborated and QualifiedName types. llvm-svn: 103517
2010-05-12 05:36:43 +08:00
if ((ClassDecl->getTagKind() == TTK_Class ||
Reinstate the fix for PR7556. A silly use of isTrivial() was suppressing copies of objects with trivial copy constructors. llvm-svn: 107857
2010-07-08 14:14:04 +08:00
ClassDecl->getTagKind() == TTK_Struct)) {
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Sequence.AddZeroInitializationStep(Entity.getType());
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
return TryConstructorInitialization(S, Entity, Kind, 0, 0, T, Sequence);
}
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
}
}
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Sequence.AddZeroInitializationStep(Entity.getType());
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
Sequence.setSequenceKind(InitializationSequence::ZeroInitialization);
}
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
/// \brief Attempt default initialization (C++ [dcl.init]p6).
static void TryDefaultInitialization(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
InitializationSequence &Sequence) {
assert(Kind.getKind() == InitializationKind::IK_Default);
// C++ [dcl.init]p6:
// To default-initialize an object of type T means:
// - if T is an array type, each element is default-initialized;
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType();
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
while (const ArrayType *Array = S.Context.getAsArrayType(DestType))
DestType = Array->getElementType();
// - if T is a (possibly cv-qualified) class type (Clause 9), the default
// constructor for T is called (and the initialization is ill-formed if
// T has no accessible default constructor);
Migrate the mish-mash of declaration checks in Sema::ActOnUninitializedDecl over to InitializationSequence (with default initialization), eliminating redundancy. More importantly, we now check that a const definition in C++ has an initilizer, which was an #if 0'd code for many, many months. A few other tweaks were needed to get everything working again: - Fix all of the places in the testsuite where we defined const objects without initializers (now that we diagnose this issue) - Teach instantiation of static data members to find the previous declaration, so that we build proper redeclaration chains. Previously, we had the redeclaration chain but built it too late to be useful, because... - Teach instantiation of static data member definitions not to try to check an initializer if a previous declaration already had an initializer. This makes sure that we don't complain about static const data members with in-class initializers and out-of-line definitions. - Move all of the incomplete-type checking logic out of Sema::FinalizeDeclaratorGroup; it makes more sense in ActOnUnitializedDecl. There may still be a few places where we can improve these diagnostics. I'll address that as a separate commit. llvm-svn: 95657
2010-02-09 15:26:29 +08:00
if (DestType->isRecordType() && S.getLangOptions().CPlusPlus) {
Relax the construction of a definition for implicit, trivial default constructors. We perform semantic checking when creating the definition, and this isn't needed in certain contexts (value initialization) but is in others (default initialization). This fixes PR7948. We add explicit code to the default initialization path to ensure the definition is both present and valid. Doug, please review. I think this follows your latest suggestion, and it ended up remarkably cleaner than I anticipated. Also let me know if similar logic should be followed for destructors and copy-constructors. llvm-svn: 111802
2010-08-23 15:55:51 +08:00
TryConstructorInitialization(S, Entity, Kind, 0, 0, DestType, Sequence);
return;
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
}
// - otherwise, no initialization is performed.
Sequence.setSequenceKind(InitializationSequence::NoInitialization);
// If a program calls for the default initialization of an object of
// a const-qualified type T, T shall be a class type with a user-provided
// default constructor.
Migrate the mish-mash of declaration checks in Sema::ActOnUninitializedDecl over to InitializationSequence (with default initialization), eliminating redundancy. More importantly, we now check that a const definition in C++ has an initilizer, which was an #if 0'd code for many, many months. A few other tweaks were needed to get everything working again: - Fix all of the places in the testsuite where we defined const objects without initializers (now that we diagnose this issue) - Teach instantiation of static data members to find the previous declaration, so that we build proper redeclaration chains. Previously, we had the redeclaration chain but built it too late to be useful, because... - Teach instantiation of static data member definitions not to try to check an initializer if a previous declaration already had an initializer. This makes sure that we don't complain about static const data members with in-class initializers and out-of-line definitions. - Move all of the incomplete-type checking logic out of Sema::FinalizeDeclaratorGroup; it makes more sense in ActOnUnitializedDecl. There may still be a few places where we can improve these diagnostics. I'll address that as a separate commit. llvm-svn: 95657
2010-02-09 15:26:29 +08:00
if (DestType.isConstQualified() && S.getLangOptions().CPlusPlus)
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
Sequence.SetFailed(InitializationSequence::FK_DefaultInitOfConst);
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
/// \brief Attempt a user-defined conversion between two types (C++ [dcl.init]),
/// which enumerates all conversion functions and performs overload resolution
/// to select the best.
static void TryUserDefinedConversion(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
Expr *Initializer,
InitializationSequence &Sequence) {
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
Sequence.setSequenceKind(InitializationSequence::UserDefinedConversion);
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType();
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
assert(!DestType->isReferenceType() && "References are handled elsewhere");
QualType SourceType = Initializer->getType();
assert((DestType->isRecordType() || SourceType->isRecordType()) &&
"Must have a class type to perform a user-defined conversion");
// Build the candidate set directly in the initialization sequence
// structure, so that it will persist if we fail.
OverloadCandidateSet &CandidateSet = Sequence.getFailedCandidateSet();
CandidateSet.clear();
// Determine whether we are allowed to call explicit constructors or
// explicit conversion operators.
bool AllowExplicit = Kind.getKind() == InitializationKind::IK_Direct;
if (const RecordType *DestRecordType = DestType->getAs<RecordType>()) {
// The type we're converting to is a class type. Enumerate its constructors
// to see if there is a suitable conversion.
CXXRecordDecl *DestRecordDecl
= cast<CXXRecordDecl>(DestRecordType->getDecl());
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
// Try to complete the type we're converting to.
if (!S.RequireCompleteType(Kind.getLocation(), DestType, 0)) {
DeclContext::lookup_iterator Con, ConEnd;
Introduce a new routine, LookupConstructors(), and use it for all constructor-name lookup. llvm-svn: 107536
2010-07-03 07:12:18 +08:00
for (llvm::tie(Con, ConEnd) = S.LookupConstructors(DestRecordDecl);
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
Con != ConEnd; ++Con) {
NamedDecl *D = *Con;
DeclAccessPair FoundDecl = DeclAccessPair::make(D, D->getAccess());
When we are performing copy initialization of a class type via its copy constructor, suppress user-defined conversions on the argument. Otherwise, we can end up in a recursion loop where the bind the argument of the copy constructor to another copy constructor call, whose argument is then a copy constructor call... Found by Boost.Regex which, alas, still isn't building. llvm-svn: 102269
2010-04-25 04:54:38 +08:00
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
// Find the constructor (which may be a template).
CXXConstructorDecl *Constructor = 0;
FunctionTemplateDecl *ConstructorTmpl
= dyn_cast<FunctionTemplateDecl>(D);
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
if (ConstructorTmpl)
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
Constructor = cast<CXXConstructorDecl>(
ConstructorTmpl->getTemplatedDecl());
When performing copy initialization via user-defined conversions, don't allow two user-defined conversions. Fixes PR6595 (again). llvm-svn: 107379
2010-07-01 11:43:00 +08:00
else
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
Constructor = cast<CXXConstructorDecl>(D);
if (!Constructor->isInvalidDecl() &&
Constructor->isConvertingConstructor(AllowExplicit)) {
if (ConstructorTmpl)
S.AddTemplateOverloadCandidate(ConstructorTmpl, FoundDecl,
/*ExplicitArgs*/ 0,
&Initializer, 1, CandidateSet,
When performing copy initialization via user-defined conversions, don't allow two user-defined conversions. Fixes PR6595 (again). llvm-svn: 107379
2010-07-01 11:43:00 +08:00
/*SuppressUserConversions=*/true);
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
else
S.AddOverloadCandidate(Constructor, FoundDecl,
&Initializer, 1, CandidateSet,
When performing copy initialization via user-defined conversions, don't allow two user-defined conversions. Fixes PR6595 (again). llvm-svn: 107379
2010-07-01 11:43:00 +08:00
/*SuppressUserConversions=*/true);
Ensure that we have completed a type before attempting initialization on that type. Fixes several problems in Boost.Interprocess. llvm-svn: 102339
2010-04-26 22:36:57 +08:00
}
}
}
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
}
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
SourceLocation DeclLoc = Initializer->getLocStart();
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
if (const RecordType *SourceRecordType = SourceType->getAs<RecordType>()) {
// The type we're converting from is a class type, enumerate its conversion
// functions.
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
Fix review comment; no visible change. llvm-svn: 91797
2009-12-21 06:12:03 +08:00
// We can only enumerate the conversion functions for a complete type; if
// the type isn't complete, simply skip this step.
if (!S.RequireCompleteType(DeclLoc, SourceType, 0)) {
CXXRecordDecl *SourceRecordDecl
= cast<CXXRecordDecl>(SourceRecordType->getDecl());
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
Give UnresolvedSet the ability to store access specifiers for each declaration. Change LookupResult to use UnresolvedSet. Also extract UnresolvedSet into its own header and make it templated over an inline capacity. llvm-svn: 93959
2010-01-20 08:46:10 +08:00
const UnresolvedSetImpl *Conversions
Fix review comment; no visible change. llvm-svn: 91797
2009-12-21 06:12:03 +08:00
= SourceRecordDecl->getVisibleConversionFunctions();
Give UnresolvedSet the ability to store access specifiers for each declaration. Change LookupResult to use UnresolvedSet. Also extract UnresolvedSet into its own header and make it templated over an inline capacity. llvm-svn: 93959
2010-01-20 08:46:10 +08:00
for (UnresolvedSetImpl::const_iterator I = Conversions->begin(),
Fix review comment; no visible change. llvm-svn: 91797
2009-12-21 06:12:03 +08:00
E = Conversions->end();
I != E; ++I) {
NamedDecl *D = *I;
CXXRecordDecl *ActingDC = cast<CXXRecordDecl>(D->getDeclContext());
if (isa<UsingShadowDecl>(D))
D = cast<UsingShadowDecl>(D)->getTargetDecl();
FunctionTemplateDecl *ConvTemplate = dyn_cast<FunctionTemplateDecl>(D);
CXXConversionDecl *Conv;
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
if (ConvTemplate)
Fix review comment; no visible change. llvm-svn: 91797
2009-12-21 06:12:03 +08:00
Conv = cast<CXXConversionDecl>(ConvTemplate->getTemplatedDecl());
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
else
Regularize support for naming conversion functions in using decls. llvm-svn: 99979
2010-03-31 09:36:47 +08:00
Conv = cast<CXXConversionDecl>(D);
Fix review comment; no visible change. llvm-svn: 91797
2009-12-21 06:12:03 +08:00
if (AllowExplicit || !Conv->isExplicit()) {
if (ConvTemplate)
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddTemplateConversionCandidate(ConvTemplate, I.getPair(),
Pass access specifiers around in overload resolution. llvm-svn: 94485
2010-01-26 09:37:31 +08:00
ActingDC, Initializer, DestType,
Fix review comment; no visible change. llvm-svn: 91797
2009-12-21 06:12:03 +08:00
CandidateSet);
else
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
S.AddConversionCandidate(Conv, I.getPair(), ActingDC,
Pass access specifiers around in overload resolution. llvm-svn: 94485
2010-01-26 09:37:31 +08:00
Initializer, DestType, CandidateSet);
Fix review comment; no visible change. llvm-svn: 91797
2009-12-21 06:12:03 +08:00
}
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
}
}
}
// Perform overload resolution. If it fails, return the failed result.
OverloadCandidateSet::iterator Best;
Introduce a specific representation for the ambiguous implicit conversion sequence. Lots of small relevant changes. Fixes some serious problems with ambiguous conversions; also possibly improves associated diagnostics. llvm-svn: 93214
2010-01-12 08:44:57 +08:00
if (OverloadingResult Result
When performing overload resolution, only compare the final conversion sequences for two conversion functions when in fact we are in the text of initialization by a user-defined conversion sequences. Fixes PR8034. llvm-svn: 113724
2010-09-12 16:07:23 +08:00
= CandidateSet.BestViableFunction(S, DeclLoc, Best, true)) {
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
Sequence.SetOverloadFailure(
InitializationSequence::FK_UserConversionOverloadFailed,
Result);
return;
}
Introduce a specific representation for the ambiguous implicit conversion sequence. Lots of small relevant changes. Fixes some serious problems with ambiguous conversions; also possibly improves associated diagnostics. llvm-svn: 93214
2010-01-12 08:44:57 +08:00
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
FunctionDecl *Function = Best->Function;
if (isa<CXXConstructorDecl>(Function)) {
// Add the user-defined conversion step. Any cv-qualification conversion is
// subsumed by the initialization.
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
Sequence.AddUserConversionStep(Function, Best->FoundDecl, DestType);
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
return;
}
// Add the user-defined conversion step that calls the conversion function.
When forming a function call or message send expression, be sure to strip cv-qualifiers from the expression's type when the language calls for it: in C, that's all the time, while C++ only does it for non-class types. Centralized the computation of the call expression type in QualType::getCallResultType() and some helper functions in other nodes (FunctionDecl, ObjCMethodDecl, FunctionType), and updated all relevant callers of getResultType() to getCallResultType(). Fixes PR7598 and PR7463, along with a bunch of getResultType() call sites that weren't stripping references off the result type (nothing stripped cv-qualifiers properly before this change). llvm-svn: 108234
2010-07-13 16:18:22 +08:00
QualType ConvType = Function->getCallResultType();
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
if (ConvType->getAs<RecordType>()) {
// If we're converting to a class type, there may be an copy if
// the resulting temporary object (possible to create an object of
// a base class type). That copy is not a separate conversion, so
// we just make a note of the actual destination type (possibly a
// base class of the type returned by the conversion function) and
// let the user-defined conversion step handle the conversion.
Sequence.AddUserConversionStep(Function, Best->FoundDecl, DestType);
return;
}
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
Sequence.AddUserConversionStep(Function, Best->FoundDecl, ConvType);
// If the conversion following the call to the conversion function
// is interesting, add it as a separate step.
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
if (Best->FinalConversion.First || Best->FinalConversion.Second ||
Best->FinalConversion.Third) {
ImplicitConversionSequence ICS;
Introduce a specific representation for the ambiguous implicit conversion sequence. Lots of small relevant changes. Fixes some serious problems with ambiguous conversions; also possibly improves associated diagnostics. llvm-svn: 93214
2010-01-12 08:44:57 +08:00
ICS.setStandard();
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
ICS.Standard = Best->FinalConversion;
Sequence.AddConversionSequenceStep(ICS, DestType);
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
InitializationSequence::InitializationSequence(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
Expr **Args,
Thread a source location into the template-argument deduction routines. There may be some other places that could take advantage of this new information, but I haven't really looked yet. llvm-svn: 95600
2010-02-09 07:07:23 +08:00
unsigned NumArgs)
: FailedCandidateSet(Kind.getLocation()) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
ASTContext &Context = S.Context;
// C++0x [dcl.init]p16:
// The semantics of initializers are as follows. The destination type is
// the type of the object or reference being initialized and the source
// type is the type of the initializer expression. The source type is not
// defined when the initializer is a braced-init-list or when it is a
// parenthesized list of expressions.
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (DestType->isDependentType() ||
Expr::hasAnyTypeDependentArguments(Args, NumArgs)) {
SequenceKind = DependentSequence;
return;
}
QualType SourceType;
Expr *Initializer = 0;
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
if (NumArgs == 1) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Initializer = Args[0];
if (!isa<InitListExpr>(Initializer))
SourceType = Initializer->getType();
}
// - If the initializer is a braced-init-list, the object is
// list-initialized (8.5.4).
if (InitListExpr *InitList = dyn_cast_or_null<InitListExpr>(Initializer)) {
TryListInitialization(S, Entity, Kind, InitList, *this);
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
return;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
// - If the destination type is a reference type, see 8.5.3.
if (DestType->isReferenceType()) {
// C++0x [dcl.init.ref]p1:
// A variable declared to be a T& or T&&, that is, "reference to type T"
// (8.3.2), shall be initialized by an object, or function, of type T or
// by an object that can be converted into a T.
// (Therefore, multiple arguments are not permitted.)
if (NumArgs != 1)
SetFailed(FK_TooManyInitsForReference);
else
TryReferenceInitialization(S, Entity, Kind, Args[0], *this);
return;
}
// - If the destination type is an array of characters, an array of
// char16_t, an array of char32_t, or an array of wchar_t, and the
// initializer is a string literal, see 8.5.2.
if (Initializer && IsStringInit(Initializer, DestType, Context)) {
TryStringLiteralInitialization(S, Entity, Kind, Initializer, *this);
return;
}
// - If the initializer is (), the object is value-initialized.
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
if (Kind.getKind() == InitializationKind::IK_Value ||
(Kind.getKind() == InitializationKind::IK_Direct && NumArgs == 0)) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
TryValueInitialization(S, Entity, Kind, *this);
return;
}
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
// Handle default initialization.
if (Kind.getKind() == InitializationKind::IK_Default){
TryDefaultInitialization(S, Entity, Kind, *this);
return;
}
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// - Otherwise, if the destination type is an array, the program is
// ill-formed.
if (const ArrayType *AT = Context.getAsArrayType(DestType)) {
if (AT->getElementType()->isAnyCharacterType())
SetFailed(FK_ArrayNeedsInitListOrStringLiteral);
else
SetFailed(FK_ArrayNeedsInitList);
return;
}
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
// Handle initialization in C
if (!S.getLangOptions().CPlusPlus) {
setSequenceKind(CAssignment);
AddCAssignmentStep(DestType);
return;
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// - If the destination type is a (possibly cv-qualified) class type:
if (DestType->isRecordType()) {
// - If the initialization is direct-initialization, or if it is
// copy-initialization where the cv-unqualified version of the
// source type is the same class as, or a derived class of, the
// class of the destination, constructors are considered. [...]
if (Kind.getKind() == InitializationKind::IK_Direct ||
(Kind.getKind() == InitializationKind::IK_Copy &&
(Context.hasSameUnqualifiedType(SourceType, DestType) ||
S.IsDerivedFrom(SourceType, DestType))))
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
TryConstructorInitialization(S, Entity, Kind, Args, NumArgs,
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Entity.getType(), *this);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// - Otherwise (i.e., for the remaining copy-initialization cases),
// user-defined conversion sequences that can convert from the source
// type to the destination type or (when a conversion function is
// used) to a derived class thereof are enumerated as described in
// 13.3.1.4, and the best one is chosen through overload resolution
// (13.3).
else
TryUserDefinedConversion(S, Entity, Kind, Initializer, *this);
return;
}
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
if (NumArgs > 1) {
SetFailed(FK_TooManyInitsForScalar);
return;
}
assert(NumArgs == 1 && "Zero-argument case handled above");
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// - Otherwise, if the source type is a (possibly cv-qualified) class
// type, conversion functions are considered.
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
if (!SourceType.isNull() && SourceType->isRecordType()) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
TryUserDefinedConversion(S, Entity, Kind, Initializer, *this);
return;
}
// - Otherwise, the initial value of the object being initialized is the
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
// (possibly converted) value of the initializer expression. Standard
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// conversions (Clause 4) will be used, if necessary, to convert the
// initializer expression to the cv-unqualified version of the
// destination type; no user-defined conversions are considered.
More refactoring. llvm-svn: 105458
2010-06-04 10:29:22 +08:00
if (S.TryImplicitConversion(*this, Entity, Initializer,
/*SuppressUserConversions*/ true,
/*AllowExplicitConversions*/ false,
/*InOverloadResolution*/ false))
SetFailed(InitializationSequence::FK_ConversionFailed);
else
setSequenceKind(StandardConversion);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
InitializationSequence::~InitializationSequence() {
for (llvm::SmallVectorImpl<Step>::iterator Step = Steps.begin(),
StepEnd = Steps.end();
Step != StepEnd; ++Step)
Step->Destroy();
}
//===----------------------------------------------------------------------===//
// Perform initialization
//===----------------------------------------------------------------------===//
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
static Sema::AssignmentAction
getAssignmentAction(const InitializedEntity &Entity) {
switch(Entity.getKind()) {
case InitializedEntity::EK_Variable:
case InitializedEntity::EK_New:
return Sema::AA_Initializing;
case InitializedEntity::EK_Parameter:
Switch the initialization of Objective-C message parameters (as occurs during message sends) over to the new initialization code and away from the C-only CheckSingleAssignmentConstraints. The enables the use of C++ types in method parameters and message arguments, as well as unifying more initialiation code overall. llvm-svn: 102035
2010-04-22 07:24:10 +08:00
if (Entity.getDecl() &&
isa<ObjCMethodDecl>(Entity.getDecl()->getDeclContext()))
return Sema::AA_Sending;
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
return Sema::AA_Passing;
case InitializedEntity::EK_Result:
return Sema::AA_Returning;
case InitializedEntity::EK_Exception:
case InitializedEntity::EK_Base:
llvm_unreachable("No assignment action for C++-specific initialization");
break;
case InitializedEntity::EK_Temporary:
// FIXME: Can we tell apart casting vs. converting?
return Sema::AA_Casting;
case InitializedEntity::EK_Member:
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
case InitializedEntity::EK_ArrayElement:
case InitializedEntity::EK_VectorElement:
Use MaybeCreateCXXExprWithTemporaries for potential destruction of created temporary. Use own initialized entity for copied in block variables. llvm-svn: 105533
2010-06-08 00:14:00 +08:00
case InitializedEntity::EK_BlockElement:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
return Sema::AA_Initializing;
}
return Sema::AA_Converting;
}
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
/// \brief Whether we should binding a created object as a temporary when
/// initializing the given entity.
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
static bool shouldBindAsTemporary(const InitializedEntity &Entity) {
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
switch (Entity.getKind()) {
Use new initialization code when dealing with [dcl.init.aggr]p12. This fixes the bug where array elements and member initializers weren't copied correctly. llvm-svn: 94340
2010-01-24 08:19:41 +08:00
case InitializedEntity::EK_ArrayElement:
case InitializedEntity::EK_Member:
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
case InitializedEntity::EK_Result:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case InitializedEntity::EK_New:
case InitializedEntity::EK_Variable:
case InitializedEntity::EK_Base:
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
case InitializedEntity::EK_VectorElement:
Add support for threadsafe statics, and make them the default (matching gcc). Daniel, I'd appreciate a review of the driver/cc1 parts. llvm-svn: 95508
2010-02-07 07:23:06 +08:00
case InitializedEntity::EK_Exception:
Use MaybeCreateCXXExprWithTemporaries for potential destruction of created temporary. Use own initialized entity for copied in block variables. llvm-svn: 105533
2010-06-08 00:14:00 +08:00
case InitializedEntity::EK_BlockElement:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
return false;
case InitializedEntity::EK_Parameter:
case InitializedEntity::EK_Temporary:
return true;
}
llvm_unreachable("missed an InitializedEntity kind?");
}
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
/// \brief Whether the given entity, when initialized with an object
/// created for that initialization, requires destruction.
static bool shouldDestroyTemporary(const InitializedEntity &Entity) {
switch (Entity.getKind()) {
case InitializedEntity::EK_Member:
case InitializedEntity::EK_Result:
case InitializedEntity::EK_New:
case InitializedEntity::EK_Base:
case InitializedEntity::EK_VectorElement:
Use MaybeCreateCXXExprWithTemporaries for potential destruction of created temporary. Use own initialized entity for copied in block variables. llvm-svn: 105533
2010-06-08 00:14:00 +08:00
case InitializedEntity::EK_BlockElement:
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
return false;
case InitializedEntity::EK_Variable:
case InitializedEntity::EK_Parameter:
case InitializedEntity::EK_Temporary:
case InitializedEntity::EK_ArrayElement:
case InitializedEntity::EK_Exception:
return true;
}
llvm_unreachable("missed an InitializedEntity kind?");
}
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
/// \brief Make a (potentially elidable) temporary copy of the object
/// provided by the given initializer by calling the appropriate copy
/// constructor.
///
/// \param S The Sema object used for type-checking.
///
/// \param T The type of the temporary object, which must either by
/// the type of the initializer expression or a superclass thereof.
///
/// \param Enter The entity being initialized.
///
/// \param CurInit The initializer expression.
///
/// \param IsExtraneousCopy Whether this is an "extraneous" copy that
/// is permitted in C++03 (but not C++0x) when binding a reference to
/// an rvalue.
///
/// \returns An expression that copies the initializer expression into
/// a temporary object, or an error expression if a copy could not be
/// created.
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
static ExprResult CopyObject(Sema &S,
When performing overload resolution, only compare the final conversion sequences for two conversion functions when in fact we are in the text of initialization by a user-defined conversion sequences. Fixes PR8034. llvm-svn: 113724
2010-09-12 16:07:23 +08:00
QualType T,
const InitializedEntity &Entity,
ExprResult CurInit,
bool IsExtraneousCopy) {
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
// Determine which class type we're copying to.
Use new initialization code when dealing with [dcl.init.aggr]p12. This fixes the bug where array elements and member initializers weren't copied correctly. llvm-svn: 94340
2010-01-24 08:19:41 +08:00
Expr *CurInitExpr = (Expr *)CurInit.get();
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
CXXRecordDecl *Class = 0;
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
if (const RecordType *Record = T->getAs<RecordType>())
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
Class = cast<CXXRecordDecl>(Record->getDecl());
if (!Class)
return move(CurInit);
// C++0x [class.copy]p34:
// When certain criteria are met, an implementation is allowed to
// omit the copy/move construction of a class object, even if the
// copy/move constructor and/or destructor for the object have
// side effects. [...]
// - when a temporary class object that has not been bound to a
// reference (12.2) would be copied/moved to a class object
// with the same cv-unqualified type, the copy/move operation
// can be omitted by constructing the temporary object
// directly into the target of the omitted copy/move
//
// Note that the other three bullets are handled elsewhere. Copy
Recognize when the named return value optimization applies in a "return" statement and mark the corresponding CXXConstructExpr as elidable. Teach CodeGen that eliding a temporary is different from eliding an object construction. This is just a baby step toward NRVO. llvm-svn: 103849
2010-05-15 08:13:29 +08:00
// elision for return statements and throw expressions are handled as part
// of constructor initialization, while copy elision for exception handlers
// is handled by the run-time.
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
bool Elidable = CurInitExpr->isTemporaryObject() &&
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
S.Context.hasSameUnqualifiedType(T, CurInitExpr->getType());
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
SourceLocation Loc;
switch (Entity.getKind()) {
case InitializedEntity::EK_Result:
Loc = Entity.getReturnLoc();
break;
case InitializedEntity::EK_Exception:
Loc = Entity.getThrowLoc();
break;
case InitializedEntity::EK_Variable:
Switch more of Sema::CheckInitializerTypes over to InitializationSequence. Specially, switch initialization of a C++ class type (either copy- or direct-initialization). Also, make sure that we create an elidable copy-construction when performing copy initialization of a C++ class variable. Fixes PR5826. llvm-svn: 91750
2009-12-19 11:01:41 +08:00
Loc = Entity.getDecl()->getLocation();
break;
Use new initialization code when dealing with [dcl.init.aggr]p12. This fixes the bug where array elements and member initializers weren't copied correctly. llvm-svn: 94340
2010-01-24 08:19:41 +08:00
case InitializedEntity::EK_ArrayElement:
case InitializedEntity::EK_Member:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case InitializedEntity::EK_Parameter:
case InitializedEntity::EK_Temporary:
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
case InitializedEntity::EK_New:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case InitializedEntity::EK_Base:
Separate EK_ArrayOrVectorElement into EK_ArrayElement and EK_VectorElement; arrays and vectors are pretty different beasts in C++. Doug, please review/comment. llvm-svn: 94279
2010-01-23 12:34:47 +08:00
case InitializedEntity::EK_VectorElement:
Use MaybeCreateCXXExprWithTemporaries for potential destruction of created temporary. Use own initialized entity for copied in block variables. llvm-svn: 105533
2010-06-08 00:14:00 +08:00
case InitializedEntity::EK_BlockElement:
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
Loc = CurInitExpr->getLocStart();
break;
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
}
When we attempt to create a temporary object of class type, be sure that the type we're copying is complete. Boost.Regex now builds, although it's failing its regression tests with our favorite "Sema doesn't consider destructor as used." assertion. llvm-svn: 102271
2010-04-25 05:09:25 +08:00
// Make sure that the type we are copying is complete.
if (S.RequireCompleteType(Loc, T, S.PDiag(diag::err_temp_copy_incomplete)))
return move(CurInit);
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
// Perform overload resolution using the class's copy constructors.
DeclContext::lookup_iterator Con, ConEnd;
Thread a source location into the template-argument deduction routines. There may be some other places that could take advantage of this new information, but I haven't really looked yet. llvm-svn: 95600
2010-02-09 07:07:23 +08:00
OverloadCandidateSet CandidateSet(Loc);
Introduce a new routine, LookupConstructors(), and use it for all constructor-name lookup. llvm-svn: 107536
2010-07-03 07:12:18 +08:00
for (llvm::tie(Con, ConEnd) = S.LookupConstructors(Class);
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
Con != ConEnd; ++Con) {
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
// Only consider copy constructors.
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
CXXConstructorDecl *Constructor = dyn_cast<CXXConstructorDecl>(*Con);
if (!Constructor || Constructor->isInvalidDecl() ||
Do not consider explicit constructors when performing a copy to a temporary object. This is blindingly obvious from reading C++ [over.match.ctor]p1, but somehow I'd missed it and it took DR152 to educate me. Adjust one test that was relying on this non-standard behavior. llvm-svn: 101688
2010-04-18 10:16:12 +08:00
!Constructor->isCopyConstructor() ||
!Constructor->isConvertingConstructor(/*AllowExplicit=*/false))
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
continue;
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
DeclAccessPair FoundDecl
= DeclAccessPair::make(Constructor, Constructor->getAccess());
S.AddOverloadCandidate(Constructor, FoundDecl,
Pass access specifiers around in overload resolution. llvm-svn: 94485
2010-01-26 09:37:31 +08:00
&CurInitExpr, 1, CandidateSet);
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
}
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
OverloadCandidateSet::iterator Best;
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
switch (CandidateSet.BestViableFunction(S, Loc, Best)) {
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case OR_Success:
break;
case OR_No_Viable_Function:
PR7245: Make binding a reference to a temporary without a usable copy constructor into an extension warning into the error that C++98 requires. llvm-svn: 105529
2010-06-07 23:58:05 +08:00
S.Diag(Loc, IsExtraneousCopy && !S.isSFINAEContext()
? diag::ext_rvalue_to_reference_temp_copy_no_viable
: diag::err_temp_copy_no_viable)
Switch more of Sema::CheckInitializerTypes over to InitializationSequence. Specially, switch initialization of a C++ class type (either copy- or direct-initialization). Also, make sure that we create an elidable copy-construction when performing copy initialization of a C++ class variable. Fixes PR5826. llvm-svn: 91750
2009-12-19 11:01:41 +08:00
<< (int)Entity.getKind() << CurInitExpr->getType()
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
<< CurInitExpr->getSourceRange();
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
CandidateSet.NoteCandidates(S, OCD_AllCandidates, &CurInitExpr, 1);
PR7245: Make binding a reference to a temporary without a usable copy constructor into an extension warning into the error that C++98 requires. llvm-svn: 105529
2010-06-07 23:58:05 +08:00
if (!IsExtraneousCopy || S.isSFINAEContext())
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
PR7245: Make binding a reference to a temporary without a usable copy constructor into an extension warning into the error that C++98 requires. llvm-svn: 105529
2010-06-07 23:58:05 +08:00
return move(CurInit);
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case OR_Ambiguous:
S.Diag(Loc, diag::err_temp_copy_ambiguous)
Switch more of Sema::CheckInitializerTypes over to InitializationSequence. Specially, switch initialization of a C++ class type (either copy- or direct-initialization). Also, make sure that we create an elidable copy-construction when performing copy initialization of a C++ class variable. Fixes PR5826. llvm-svn: 91750
2009-12-19 11:01:41 +08:00
<< (int)Entity.getKind() << CurInitExpr->getType()
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
<< CurInitExpr->getSourceRange();
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
CandidateSet.NoteCandidates(S, OCD_ViableCandidates, &CurInitExpr, 1);
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case OR_Deleted:
S.Diag(Loc, diag::err_temp_copy_deleted)
Switch more of Sema::CheckInitializerTypes over to InitializationSequence. Specially, switch initialization of a C++ class type (either copy- or direct-initialization). Also, make sure that we create an elidable copy-construction when performing copy initialization of a C++ class variable. Fixes PR5826. llvm-svn: 91750
2009-12-19 11:01:41 +08:00
<< (int)Entity.getKind() << CurInitExpr->getType()
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
<< CurInitExpr->getSourceRange();
S.Diag(Best->Function->getLocation(), diag::note_unavailable_here)
<< Best->Function->isDeleted();
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
}
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
CXXConstructorDecl *Constructor = cast<CXXConstructorDecl>(Best->Function);
Sundry incremental steps towards killing off Action. llvm-svn: 111795
2010-08-23 14:44:23 +08:00
ASTOwningVector<Expr*> ConstructorArgs(S);
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
CurInit.release(); // Ownership transferred into MultiExprArg, below.
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
Pass the InitializedEntity to Sema::CheckConstructorAccess and use it to report different diagnostics depending on which entity is being initialized. llvm-svn: 102010
2010-04-22 02:47:17 +08:00
S.CheckConstructorAccess(Loc, Constructor, Entity,
PR7245: Make binding a reference to a temporary without a usable copy constructor into an extension warning into the error that C++98 requires. llvm-svn: 105529
2010-06-07 23:58:05 +08:00
Best->FoundDecl.getAccess(), IsExtraneousCopy);
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
if (IsExtraneousCopy) {
// If this is a totally extraneous copy for C++03 reference
// binding purposes, just return the original initialization
When checking the copy constructor for the optional copy during a reference binding to an rvalue of reference-compatible type, check parameters after the first for complete parameter types and build any required default function arguments. We're effectively simulating the type-checking for a call without building the call itself. llvm-svn: 101705
2010-04-18 15:57:34 +08:00
// expression. We don't generate an (elided) copy operation here
// because doing so would require us to pass down a flag to avoid
// infinite recursion, where each step adds another extraneous,
// elidable copy.
// Instantiate the default arguments of any extra parameters in
// the selected copy constructor, as if we were going to create a
// proper call to the copy constructor.
for (unsigned I = 1, N = Constructor->getNumParams(); I != N; ++I) {
ParmVarDecl *Parm = Constructor->getParamDecl(I);
if (S.RequireCompleteType(Loc, Parm->getType(),
S.PDiag(diag::err_call_incomplete_argument)))
break;
// Build the default argument expression; we don't actually care
// if this succeeds or not, because this routine will complain
// if there was a problem.
S.BuildCXXDefaultArgExpr(Loc, Constructor, Parm);
}
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
return S.Owned(CurInitExpr);
}
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
// Determine the arguments required to actually perform the
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
// constructor call (we might have derived-to-base conversions, or
// the copy constructor may have default arguments).
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
if (S.CompleteConstructorCall(Constructor, MultiExprArg(&CurInitExpr, 1),
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
Loc, ConstructorArgs))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Improve our handling of user-defined conversions as part of overload resolution. There are two sources of problems involving user-defined conversions that this change eliminates, along with providing simpler interfaces for checking implicit conversions: - It eliminates a case of infinite recursion found in Boost. - It eliminates the search for the constructor needed to copy a temporary generated by an implicit conversion from overload resolution. Overload resolution assumes that, if it gets a value of the parameter's class type (or a derived class thereof), there is a way to copy if... even if there isn't. We now model this properly. llvm-svn: 101680
2010-04-18 06:01:05 +08:00
When copying a temporary object to initialize an entity for which the temporary needs to be bound, bind the copy object. Otherwise, we won't end up calling the destructor for the copy. Fixes Boost.Optional. llvm-svn: 102290
2010-04-25 08:55:24 +08:00
// Actually perform the constructor call.
CurInit = S.BuildCXXConstructExpr(Loc, T, Constructor, Elidable,
Remove a header dependency from Sema.h at the cost of some type safety. If someone wants to fix this some other way.... llvm-svn: 111905
2010-08-24 15:32:53 +08:00
move_arg(ConstructorArgs),
/*ZeroInit*/ false,
CXXConstructExpr::CK_Complete);
When copying a temporary object to initialize an entity for which the temporary needs to be bound, bind the copy object. Otherwise, we won't end up calling the destructor for the copy. Fixes Boost.Optional. llvm-svn: 102290
2010-04-25 08:55:24 +08:00
// If we're supposed to bind temporaries, do so.
if (!CurInit.isInvalid() && shouldBindAsTemporary(Entity))
CurInit = S.MaybeBindToTemporary(CurInit.takeAs<Expr>());
return move(CurInit);
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Whenever we complain about a failed initialization of a function or method parameter, provide a note pointing at the parameter itself so the user does not have to manually look for the function/method being called and match up parameters to arguments. For example, we now get: t.c:4:5: warning: incompatible pointer types passing 'long *' to parameter of type 'int *' [-pedantic] f(long_ptr); ^~~~~~~~ t.c:1:13: note: passing argument to parameter 'x' here void f(int *x); ^ llvm-svn: 102038
2010-04-22 08:20:18 +08:00
void InitializationSequence::PrintInitLocationNote(Sema &S,
const InitializedEntity &Entity) {
if (Entity.getKind() == InitializedEntity::EK_Parameter && Entity.getDecl()) {
if (Entity.getDecl()->getLocation().isInvalid())
return;
if (Entity.getDecl()->getDeclName())
S.Diag(Entity.getDecl()->getLocation(), diag::note_parameter_named_here)
<< Entity.getDecl()->getDeclName();
else
S.Diag(Entity.getDecl()->getLocation(), diag::note_parameter_here);
}
}
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
InitializationSequence::Perform(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
MultiExprArg Args,
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
QualType *ResultType) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (SequenceKind == FailedSequence) {
unsigned NumArgs = Args.size();
Diagnose(S, Entity, Kind, (Expr **)Args.release(), NumArgs);
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
if (SequenceKind == DependentSequence) {
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
// If the declaration is a non-dependent, incomplete array type
// that has an initializer, then its type will be completed once
// the initializer is instantiated.
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
if (ResultType && !Entity.getType()->isDependentType() &&
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
Args.size() == 1) {
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DeclType = Entity.getType();
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
if (const IncompleteArrayType *ArrayT
= S.Context.getAsIncompleteArrayType(DeclType)) {
// FIXME: We don't currently have the ability to accurately
// compute the length of an initializer list without
// performing full type-checking of the initializer list
// (since we have to determine where braces are implicitly
// introduced and such). So, we fall back to making the array
// type a dependently-sized array type with no specified
// bound.
if (isa<InitListExpr>((Expr *)Args.get()[0])) {
SourceRange Brackets;
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
// Scavange the location of the brackets from the entity, if we can.
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
if (DeclaratorDecl *DD = Entity.getDecl()) {
if (TypeSourceInfo *TInfo = DD->getTypeSourceInfo()) {
TypeLoc TL = TInfo->getTypeLoc();
if (IncompleteArrayTypeLoc *ArrayLoc
= dyn_cast<IncompleteArrayTypeLoc>(&TL))
Brackets = ArrayLoc->getBracketsRange();
}
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
}
*ResultType
= S.Context.getDependentSizedArrayType(ArrayT->getElementType(),
/*NumElts=*/0,
ArrayT->getSizeModifier(),
ArrayT->getIndexTypeCVRQualifiers(),
Brackets);
}
}
}
Switch compound literals over to InitializationSequence. llvm-svn: 91882
2009-12-22 10:35:53 +08:00
if (Kind.getKind() == InitializationKind::IK_Copy || Kind.isExplicitCast())
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
return ExprResult(Args.release()[0]);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
A dependent initializer with zero arguments should return a NULL initializer (for no initialization) rather than a ParenListExpr with zero arguments in it. llvm-svn: 95382
2010-02-05 15:56:11 +08:00
if (Args.size() == 0)
return S.Owned((Expr *)0);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
unsigned NumArgs = Args.size();
return S.Owned(new (S.Context) ParenListExpr(S.Context,
SourceLocation(),
(Expr **)Args.release(),
NumArgs,
SourceLocation()));
}
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
if (SequenceKind == NoInitialization)
return S.Owned((Expr *)0);
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType().getNonReferenceType();
// FIXME: Ugly hack around the fact that Entity.getType() is not
Switch file-scope assignment initialization over to InitializationSequence. llvm-svn: 91881
2009-12-22 10:10:53 +08:00
// the same as Entity.getDecl()->getType() in cases involving type merging,
// and we want latter when it makes sense.
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
if (ResultType)
Switch file-scope assignment initialization over to InitializationSequence. llvm-svn: 91881
2009-12-22 10:10:53 +08:00
*ResultType = Entity.getDecl() ? Entity.getDecl()->getType() :
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
Entity.getType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult CurInit = S.Owned((Expr *)0);
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
assert(!Steps.empty() && "Cannot have an empty initialization sequence");
// For initialization steps that start with a single initializer,
// grab the only argument out the Args and place it into the "current"
// initializer.
switch (Steps.front().Kind) {
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case SK_ResolveAddressOfOverloadedFunction:
case SK_CastDerivedToBaseRValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_CastDerivedToBaseXValue:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case SK_CastDerivedToBaseLValue:
case SK_BindReference:
case SK_BindReferenceToTemporary:
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
case SK_ExtraneousCopyToTemporary:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case SK_UserConversion:
case SK_QualificationConversionLValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_QualificationConversionXValue:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case SK_QualificationConversionRValue:
case SK_ConversionSequence:
case SK_ListInitialization:
case SK_CAssignment:
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
case SK_StringInit:
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
case SK_ObjCObjectConversion:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
assert(Args.size() == 1);
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
CurInit = ExprResult(((Expr **)(Args.get()))[0]->Retain());
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
if (CurInit.isInvalid())
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
break;
case SK_ConstructorInitialization:
case SK_ZeroInitialization:
break;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
// Walk through the computed steps for the initialization sequence,
// performing the specified conversions along the way.
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
bool ConstructorInitRequiresZeroInit = false;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
for (step_iterator Step = step_begin(), StepEnd = step_end();
Step != StepEnd; ++Step) {
if (CurInit.isInvalid())
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Expr *CurInitExpr = (Expr *)CurInit.get();
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
QualType SourceType = CurInitExpr? CurInitExpr->getType() : QualType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
switch (Step->Kind) {
case SK_ResolveAddressOfOverloadedFunction:
// Overload resolution determined which function invoke; update the
// initializer to reflect that choice.
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
S.CheckAddressOfMemberAccess(CurInitExpr, Step->Function.FoundDecl);
Diagnose deprecated/unavailable functions selected by overload resolution. Fixes rdar://problem/4232969, or at least the clang parts of it. llvm-svn: 103191
2010-05-07 02:15:07 +08:00
S.DiagnoseUseOfDecl(Step->Function.FoundDecl, Kind.getLocation());
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
CurInit = S.FixOverloadedFunctionReference(move(CurInit),
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
Step->Function.FoundDecl,
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
Step->Function.Function);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
case SK_CastDerivedToBaseRValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_CastDerivedToBaseXValue:
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case SK_CastDerivedToBaseLValue: {
// We have a derived-to-base cast that produces either an rvalue or an
// lvalue. Perform that cast.
Store inheritance paths after CastExprs instead of inside them. This takes some trickery since CastExpr has subclasses (and indeed, is abstract). Also, smoosh the CastKind into the bitfield from Expr. Drops two words of storage from Expr in the common case of expressions which don't need inheritance paths. Avoids a separate allocation and another word of overhead in cases needing inheritance paths. Also has the advantage of not leaking memory, since destructors for AST nodes are never run. llvm-svn: 110507
2010-08-07 14:22:56 +08:00
CXXCastPath BasePath;
Actually produce base paths for CastExprs of kind CK_DerivedToBase. llvm-svn: 102259
2010-04-25 03:06:50 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Casts to inaccessible base classes are allowed with C-style casts.
bool IgnoreBaseAccess = Kind.isCStyleOrFunctionalCast();
if (S.CheckDerivedToBaseConversion(SourceType, Step->Type,
CurInitExpr->getLocStart(),
Actually produce base paths for CastExprs of kind CK_DerivedToBase. llvm-svn: 102259
2010-04-25 03:06:50 +08:00
CurInitExpr->getSourceRange(),
&BasePath, IgnoreBaseAccess))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Rework when and how vtables are emitted, by tracking where vtables are "used" (e.g., we will refer to the vtable in the generated code) and when they are defined (i.e., because we've seen the key function definition). Previously, we were effectively tracking "potential definitions" rather than uses, so we were a bit too eager about emitting vtables for classes without key functions. The new scheme: - For every use of a vtable, Sema calls MarkVTableUsed() to indicate the use. For example, this occurs when calling a virtual member function of the class, defining a constructor of that class type, dynamic_cast'ing from that type to a derived class, casting to/through a virtual base class, etc. - For every definition of a vtable, Sema calls MarkVTableUsed() to indicate the definition. This happens at the end of the translation unit for classes whose key function has been defined (so we can delay computation of the key function; see PR6564), and will also occur with explicit template instantiation definitions. - For every vtable defined/used, we mark all of the virtual member functions of that vtable as defined/used, unless we know that the key function is in another translation unit. This instantiates virtual member functions when needed. - At the end of the translation unit, Sema tells CodeGen (via the ASTConsumer) which vtables must be defined (CodeGen will define them) and which may be used (for which CodeGen will define the vtables lazily). From a language perspective, both the old and the new schemes are permissible: we're allowed to instantiate virtual member functions whenever we want per the standard. However, all other C++ compilers were more lazy than we were, and our eagerness was both a performance issue (we instantiated too much) and a portability problem (we broke Boost test cases, which now pass). Notes: (1) There's a ton of churn in the tests, because the order in which vtables get emitted to IR has changed. I've tried to isolate some of the larger tests from these issues. (2) Some diagnostics related to implicitly-instantiated/implicitly-defined virtual member functions have moved to the point of first use/definition. It's better this way. (3) I could use a review of the places where we MarkVTableUsed, to see if I missed any place where the language effectively requires a vtable. Fixes PR7114 and PR6564. llvm-svn: 103718
2010-05-14 00:44:06 +08:00
if (S.BasePathInvolvesVirtualBase(BasePath)) {
QualType T = SourceType;
if (const PointerType *Pointer = T->getAs<PointerType>())
T = Pointer->getPointeeType();
if (const RecordType *RecordTy = T->getAs<RecordType>())
S.MarkVTableUsed(CurInitExpr->getLocStart(),
cast<CXXRecordDecl>(RecordTy->getDecl()));
}
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
ExprValueKind VK =
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
Step->Kind == SK_CastDerivedToBaseLValue ?
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK_LValue :
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
(Step->Kind == SK_CastDerivedToBaseXValue ?
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK_XValue :
VK_RValue);
Store inheritance paths after CastExprs instead of inside them. This takes some trickery since CastExpr has subclasses (and indeed, is abstract). Also, smoosh the CastKind into the bitfield from Expr. Drops two words of storage from Expr in the common case of expressions which don't need inheritance paths. Avoids a separate allocation and another word of overhead in cases needing inheritance paths. Also has the advantage of not leaking memory, since destructors for AST nodes are never run. llvm-svn: 110507
2010-08-07 14:22:56 +08:00
CurInit = S.Owned(ImplicitCastExpr::Create(S.Context,
Step->Type,
GCC didn't care for my attempt at API compatibility, so brute-force everything to the new constants. llvm-svn: 112047
2010-08-25 19:45:40 +08:00
CK_DerivedToBase,
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
CurInit.get(),
&BasePath, VK));
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
}
case SK_BindReference:
if (FieldDecl *BitField = CurInitExpr->getBitField()) {
// References cannot bind to bit fields (C++ [dcl.init.ref]p5).
S.Diag(Kind.getLocation(), diag::err_reference_bind_to_bitfield)
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
<< Entity.getType().isVolatileQualified()
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
<< BitField->getDeclName()
<< CurInitExpr->getSourceRange();
S.Diag(BitField->getLocation(), diag::note_bitfield_decl);
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
Simplify InitListChecker::CheckReferenceType llvm-svn: 94793
2010-01-29 10:47:33 +08:00
Diagnose binding a non-const reference to a vector element. llvm-svn: 94963
2010-02-01 01:18:49 +08:00
if (CurInitExpr->refersToVectorElement()) {
Fix this comment. llvm-svn: 95104
2010-02-03 03:02:38 +08:00
// References cannot bind to vector elements.
Diagnose binding a non-const reference to a vector element. llvm-svn: 94963
2010-02-01 01:18:49 +08:00
S.Diag(Kind.getLocation(), diag::err_reference_bind_to_vector_element)
<< Entity.getType().isVolatileQualified()
<< CurInitExpr->getSourceRange();
Whenever we complain about a failed initialization of a function or method parameter, provide a note pointing at the parameter itself so the user does not have to manually look for the function/method being called and match up parameters to arguments. For example, we now get: t.c:4:5: warning: incompatible pointer types passing 'long *' to parameter of type 'int *' [-pedantic] f(long_ptr); ^~~~~~~~ t.c:1:13: note: passing argument to parameter 'x' here void f(int *x); ^ llvm-svn: 102038
2010-04-22 08:20:18 +08:00
PrintInitLocationNote(S, Entity);
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Diagnose binding a non-const reference to a vector element. llvm-svn: 94963
2010-02-01 01:18:49 +08:00
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Reference binding does not have any corresponding ASTs.
// Check exception specifications
if (S.CheckExceptionSpecCompatibility(CurInitExpr, DestType))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Start creating CXXBindReferenceExpr nodes when binding complex types to references. llvm-svn: 94964
2010-02-01 02:34:51 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
Start creating CXXBindReferenceExpr nodes when binding complex types to references. llvm-svn: 94964
2010-02-01 02:34:51 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case SK_BindReferenceToTemporary:
Revert the new reference binding code; I came up with a way simpler solution for the reference binding bug that is preventing self-hosting. llvm-svn: 95223
2010-02-04 00:38:03 +08:00
// Reference binding does not have any corresponding ASTs.
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Check exception specifications
if (S.CheckExceptionSpecCompatibility(CurInitExpr, DestType))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
case SK_ExtraneousCopyToTemporary:
CurInit = CopyObject(S, Step->Type, Entity, move(CurInit),
/*IsExtraneousCopy=*/true);
break;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case SK_UserConversion: {
// We have a user-defined conversion that invokes either a constructor
// or a conversion function.
GCC didn't care for my attempt at API compatibility, so brute-force everything to the new constants. llvm-svn: 112047
2010-08-25 19:45:40 +08:00
CastKind CastKind = CK_Unknown;
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
bool IsCopy = false;
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
FunctionDecl *Fn = Step->Function.Function;
DeclAccessPair FoundFn = Step->Function.FoundDecl;
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
bool CreatedObject = false;
Kill off two more uses of Sema::CheckReferenceInit in favor of the new initialization code. Exposed a bug where we were not marking an implicit conversion as an lvalue when we were forming a call to a conversion function whose return type is a reference. llvm-svn: 99459
2010-03-25 08:20:38 +08:00
bool IsLvalue = false;
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
if (CXXConstructorDecl *Constructor = dyn_cast<CXXConstructorDecl>(Fn)) {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Build a call to the selected constructor.
Sundry incremental steps towards killing off Action. llvm-svn: 111795
2010-08-23 14:44:23 +08:00
ASTOwningVector<Expr*> ConstructorArgs(S);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
SourceLocation Loc = CurInitExpr->getLocStart();
CurInit.release(); // Ownership transferred into MultiExprArg, below.
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Determine the arguments required to actually perform the constructor
// call.
if (S.CompleteConstructorCall(Constructor,
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
MultiExprArg(&CurInitExpr, 1),
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
Loc, ConstructorArgs))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Build the an expression that constructs a temporary.
CurInit = S.BuildCXXConstructExpr(Loc, Step->Type, Constructor,
Remove a header dependency from Sema.h at the cost of some type safety. If someone wants to fix this some other way.... llvm-svn: 111905
2010-08-24 15:32:53 +08:00
move_arg(ConstructorArgs),
/*ZeroInit*/ false,
CXXConstructExpr::CK_Complete);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (CurInit.isInvalid())
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
Pass the InitializedEntity to Sema::CheckConstructorAccess and use it to report different diagnostics depending on which entity is being initialized. llvm-svn: 102010
2010-04-22 02:47:17 +08:00
S.CheckConstructorAccess(Kind.getLocation(), Constructor, Entity,
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
FoundFn.getAccess());
Diagnose deprecated/unavailable functions selected by overload resolution. Fixes rdar://problem/4232969, or at least the clang parts of it. llvm-svn: 103191
2010-05-07 02:15:07 +08:00
S.DiagnoseUseOfDecl(FoundFn, Kind.getLocation());
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
GCC didn't care for my attempt at API compatibility, so brute-force everything to the new constants. llvm-svn: 112047
2010-08-25 19:45:40 +08:00
CastKind = CK_ConstructorConversion;
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
QualType Class = S.Context.getTypeDeclType(Constructor->getParent());
if (S.Context.hasSameUnqualifiedType(SourceType, Class) ||
S.IsDerivedFrom(SourceType, Class))
IsCopy = true;
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
CreatedObject = true;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
} else {
// Build a call to the conversion function.
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
CXXConversionDecl *Conversion = cast<CXXConversionDecl>(Fn);
Kill off two more uses of Sema::CheckReferenceInit in favor of the new initialization code. Exposed a bug where we were not marking an implicit conversion as an lvalue when we were forming a call to a conversion function whose return type is a reference. llvm-svn: 99459
2010-03-25 08:20:38 +08:00
IsLvalue = Conversion->getResultType()->isLValueReferenceType();
Perform access control for the implicit base and member destructor calls required when emitting a destructor definition. llvm-svn: 98609
2010-03-16 13:22:47 +08:00
S.CheckMemberOperatorAccess(Kind.getLocation(), CurInitExpr, 0,
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
FoundFn);
Diagnose deprecated/unavailable functions selected by overload resolution. Fixes rdar://problem/4232969, or at least the clang parts of it. llvm-svn: 103191
2010-05-07 02:15:07 +08:00
S.DiagnoseUseOfDecl(FoundFn, Kind.getLocation());
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// FIXME: Should we move this initialization into a separate
// derived-to-base conversion? I believe the answer is "no", because
// we don't want to turn off access control here for c-style casts.
Reinstate r97674 with a fix for the assertion that was firing in <list> llvm-svn: 97686
2010-03-04 07:55:11 +08:00
if (S.PerformObjectArgumentInitialization(CurInitExpr, /*Qualifier=*/0,
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
FoundFn, Conversion))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Do a little dance to make sure that CurInit has the proper
// pointer.
CurInit.release();
// Build the actual call to the conversion function.
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
CurInit = S.Owned(S.BuildCXXMemberCallExpr(CurInitExpr, FoundFn,
Conversion));
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (CurInit.isInvalid() || !CurInit.get())
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
GCC didn't care for my attempt at API compatibility, so brute-force everything to the new constants. llvm-svn: 112047
2010-08-25 19:45:40 +08:00
CastKind = CK_UserDefinedConversion;
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
CreatedObject = Conversion->getResultType()->isRecordType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
bool RequiresCopy = !IsCopy &&
getKind() != InitializationSequence::ReferenceBinding;
if (RequiresCopy || shouldBindAsTemporary(Entity))
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
CurInit = S.MaybeBindToTemporary(CurInit.takeAs<Expr>());
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
else if (CreatedObject && shouldDestroyTemporary(Entity)) {
CurInitExpr = static_cast<Expr *>(CurInit.get());
QualType T = CurInitExpr->getType();
if (const RecordType *Record = T->getAs<RecordType>()) {
Add a new routine Sema::LookupDestructor and make all destructor-lookup calls use that routine llvm-svn: 107444
2010-07-02 06:47:18 +08:00
CXXDestructorDecl *Destructor
= S.LookupDestructor(cast<CXXRecordDecl>(Record->getDecl()));
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
S.CheckDestructorAccess(CurInitExpr->getLocStart(), Destructor,
S.PDiag(diag::err_access_dtor_temp) << T);
S.MarkDeclarationReferenced(CurInitExpr->getLocStart(), Destructor);
}
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
CurInitExpr = CurInit.takeAs<Expr>();
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
// FIXME: xvalues
Store inheritance paths after CastExprs instead of inside them. This takes some trickery since CastExpr has subclasses (and indeed, is abstract). Also, smoosh the CastKind into the bitfield from Expr. Drops two words of storage from Expr in the common case of expressions which don't need inheritance paths. Avoids a separate allocation and another word of overhead in cases needing inheritance paths. Also has the advantage of not leaking memory, since destructors for AST nodes are never run. llvm-svn: 110507
2010-08-07 14:22:56 +08:00
CurInit = S.Owned(ImplicitCastExpr::Create(S.Context,
CurInitExpr->getType(),
CastKind, CurInitExpr, 0,
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
IsLvalue ? VK_LValue : VK_RValue));
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
if (RequiresCopy)
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
CurInit = CopyObject(S, Entity.getType().getNonReferenceType(), Entity,
move(CurInit), /*IsExtraneousCopy=*/false);
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
}
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case SK_QualificationConversionLValue:
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_QualificationConversionXValue:
case SK_QualificationConversionRValue: {
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
// Perform a qualification conversion; these can never go wrong.
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
ExprValueKind VK =
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
Step->Kind == SK_QualificationConversionLValue ?
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK_LValue :
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
(Step->Kind == SK_QualificationConversionXValue ?
More incremental progress towards not including Expr.h in Sema.h. llvm-svn: 112044
2010-08-25 18:28:54 +08:00
VK_XValue :
VK_RValue);
GCC didn't care for my attempt at API compatibility, so brute-force everything to the new constants. llvm-svn: 112047
2010-08-25 19:45:40 +08:00
S.ImpCastExprToType(CurInitExpr, Step->Type, CK_NoOp, VK);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
CurInit.release();
CurInit = S.Owned(CurInitExpr);
break;
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
}
Switch the checking of implicit casts for static_cast, C-style, and functional casts over to InitializationSequence, eliminating a caller of Sema::TryImplicitConversion. We also get access and ambiguity checking "for free". More cleanups to come in this routine. llvm-svn: 101526
2010-04-17 03:30:02 +08:00
case SK_ConversionSequence: {
bool IgnoreBaseAccess = Kind.isCStyleOrFunctionalCast();
if (S.PerformImplicitConversion(CurInitExpr, Step->Type, *Step->ICS,
Sema::AA_Converting, IgnoreBaseAccess))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
CurInit.release();
Switch the checking of implicit casts for static_cast, C-style, and functional casts over to InitializationSequence, eliminating a caller of Sema::TryImplicitConversion. We also get access and ambiguity checking "for free". More cleanups to come in this routine. llvm-svn: 101526
2010-04-17 03:30:02 +08:00
CurInit = S.Owned(CurInitExpr);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
Switch the checking of implicit casts for static_cast, C-style, and functional casts over to InitializationSequence, eliminating a caller of Sema::TryImplicitConversion. We also get access and ambiguity checking "for free". More cleanups to come in this routine. llvm-svn: 101526
2010-04-17 03:30:02 +08:00
}
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
case SK_ListInitialization: {
InitListExpr *InitList = cast<InitListExpr>(CurInitExpr);
QualType Ty = Step->Type;
Eliminate Sema::CheckValueInitialization; its callers now use InitializationSequence to perform the actual initialization. Also, introduced the notion of a tree of initialized entities, so that we can know where an initialization began when dealing with nested initializations (as occur when performing list initialization). This will, eventually, be useful for producing better diagnostics when list initialization fails, because we can show the path from the top-level object being initialized down to the actual subobject where initialization failed. llvm-svn: 91516
2009-12-16 14:35:08 +08:00
if (S.CheckInitList(Entity, InitList, ResultType? *ResultType : Ty))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
CurInit.release();
CurInit = S.Owned(InitList);
break;
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
case SK_ConstructorInitialization: {
Collapse the three separate initialization paths in TryStaticImplicitCast (for references, class types, and everything else, respectively) into a single invocation of InitializationSequence. One of the paths (for class types) was the only client of Sema::TryInitializationByConstructor, which I have eliminated. This also simplified the interface for much of the cast-checking logic, eliminating yet more code. I've kept the representation of C++ functional casts with <> 1 arguments the same, despite the fact that I hate it. That fix will come soon. To satisfy my paranoia, I've bootstrapped + tested Clang with these changes. llvm-svn: 101549
2010-04-17 06:09:46 +08:00
unsigned NumArgs = Args.size();
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
CXXConstructorDecl *Constructor
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
= cast<CXXConstructorDecl>(Step->Function.Function);
Recognize when the named return value optimization applies in a "return" statement and mark the corresponding CXXConstructExpr as elidable. Teach CodeGen that eliding a temporary is different from eliding an object construction. This is just a baby step toward NRVO. llvm-svn: 103849
2010-05-15 08:13:29 +08:00
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
// Build a call to the selected constructor.
Sundry incremental steps towards killing off Action. llvm-svn: 111795
2010-08-23 14:44:23 +08:00
ASTOwningVector<Expr*> ConstructorArgs(S);
Added extra check when looking for location of '=' in a copy initialization. llvm-svn: 109025
2010-07-22 02:40:47 +08:00
SourceLocation Loc = (Kind.isCopyInit() && Kind.getEqualLoc().isValid())
? Kind.getEqualLoc()
: Kind.getLocation();
Relax the construction of a definition for implicit, trivial default constructors. We perform semantic checking when creating the definition, and this isn't needed in certain contexts (value initialization) but is in others (default initialization). This fixes PR7948. We add explicit code to the default initialization path to ensure the definition is both present and valid. Doug, please review. I think this follows your latest suggestion, and it ended up remarkably cleaner than I anticipated. Also let me know if similar logic should be followed for destructors and copy-constructors. llvm-svn: 111802
2010-08-23 15:55:51 +08:00
if (Kind.getKind() == InitializationKind::IK_Default) {
// Force even a trivial, implicit default constructor to be
// semantically checked. We do this explicitly because we don't build
// the definition for completely trivial constructors.
CXXRecordDecl *ClassDecl = Constructor->getParent();
assert(ClassDecl && "No parent class for constructor.");
if (Constructor->isImplicit() && Constructor->isDefaultConstructor() &&
ClassDecl->hasTrivialConstructor() && !Constructor->isUsed(false))
S.DefineImplicitDefaultConstructor(Loc, Constructor);
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
// Determine the arguments required to actually perform the constructor
// call.
if (S.CompleteConstructorCall(Constructor, move(Args),
Loc, ConstructorArgs))
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
Relax the construction of a definition for implicit, trivial default constructors. We perform semantic checking when creating the definition, and this isn't needed in certain contexts (value initialization) but is in others (default initialization). This fixes PR7948. We add explicit code to the default initialization path to ensure the definition is both present and valid. Doug, please review. I think this follows your latest suggestion, and it ended up remarkably cleaner than I anticipated. Also let me know if similar logic should be followed for destructors and copy-constructors. llvm-svn: 111802
2010-08-23 15:55:51 +08:00
Use CXXTemporaryObjectExpr for explicitly-constructed temporaries. We used to do this, but it got lost when we switched functional-style cast syntax over to using the new initialization code. Fixes PR6457. llvm-svn: 97568
2010-03-03 01:18:33 +08:00
if (Entity.getKind() == InitializedEntity::EK_Temporary &&
Collapse the three separate initialization paths in TryStaticImplicitCast (for references, class types, and everything else, respectively) into a single invocation of InitializationSequence. One of the paths (for class types) was the only client of Sema::TryInitializationByConstructor, which I have eliminated. This also simplified the interface for much of the cast-checking logic, eliminating yet more code. I've kept the representation of C++ functional casts with <> 1 arguments the same, despite the fact that I hate it. That fix will come soon. To satisfy my paranoia, I've bootstrapped + tested Clang with these changes. llvm-svn: 101549
2010-04-17 06:09:46 +08:00
NumArgs != 1 && // FIXME: Hack to work around cast weirdness
Use CXXTemporaryObjectExpr for explicitly-constructed temporaries. We used to do this, but it got lost when we switched functional-style cast syntax over to using the new initialization code. Fixes PR6457. llvm-svn: 97568
2010-03-03 01:18:33 +08:00
(Kind.getKind() == InitializationKind::IK_Direct ||
Kind.getKind() == InitializationKind::IK_Value)) {
// An explicitly-constructed temporary, e.g., X(1, 2).
unsigned NumExprs = ConstructorArgs.size();
Expr **Exprs = (Expr **)ConstructorArgs.take();
Fix source location of the initializer in a copy initialization. Back out hack in objc rewriter. fixes radar 8213998. llvm-svn: 109024
2010-07-22 02:31:47 +08:00
S.MarkDeclarationReferenced(Loc, Constructor);
Provide proper type-source location information for CXXTemporaryObjectExpr, CXXScalarValueInitExpr, and CXXUnresolvedConstructExpr, getting rid of a bunch of FIXMEs in the process. llvm-svn: 113319
2010-09-08 08:15:04 +08:00
TypeSourceInfo *TSInfo = Entity.getTypeSourceInfo();
if (!TSInfo)
TSInfo = S.Context.getTrivialTypeSourceInfo(Entity.getType(), Loc);
Use CXXTemporaryObjectExpr for explicitly-constructed temporaries. We used to do this, but it got lost when we switched functional-style cast syntax over to using the new initialization code. Fixes PR6457. llvm-svn: 97568
2010-03-03 01:18:33 +08:00
CurInit = S.Owned(new (S.Context) CXXTemporaryObjectExpr(S.Context,
Constructor,
Provide proper type-source location information for CXXTemporaryObjectExpr, CXXScalarValueInitExpr, and CXXUnresolvedConstructExpr, getting rid of a bunch of FIXMEs in the process. llvm-svn: 113319
2010-09-08 08:15:04 +08:00
TSInfo,
Use CXXTemporaryObjectExpr for explicitly-constructed temporaries. We used to do this, but it got lost when we switched functional-style cast syntax over to using the new initialization code. Fixes PR6457. llvm-svn: 97568
2010-03-03 01:18:33 +08:00
Exprs,
NumExprs,
When explicitly building a temporary object (CXXTemporaryObjectExpr), keep track of whether we need to zero-initialize storage prior to calling its constructor. Previously, we were only tracking this when implicitly constructing the object (a CXXConstructExpr). Fixes Boost's value-initialization tests, which means that the Boost.Config library now passes all of its tests. llvm-svn: 102461
2010-04-28 04:36:09 +08:00
Kind.getParenRange().getEnd(),
ConstructorInitRequiresZeroInit));
Add an enum to CXXConstructExpr so we can determine if the construction expression constructs a non-virtual or virtual base. llvm-svn: 102879
2010-05-03 06:54:08 +08:00
} else {
CXXConstructExpr::ConstructionKind ConstructKind =
CXXConstructExpr::CK_Complete;
if (Entity.getKind() == InitializedEntity::EK_Base) {
ConstructKind = Entity.getBaseSpecifier()->isVirtual() ?
CXXConstructExpr::CK_VirtualBase :
CXXConstructExpr::CK_NonVirtualBase;
}
Recognize when the named return value optimization applies in a "return" statement and mark the corresponding CXXConstructExpr as elidable. Teach CodeGen that eliding a temporary is different from eliding an object construction. This is just a baby step toward NRVO. llvm-svn: 103849
2010-05-15 08:13:29 +08:00
// If the entity allows NRVO, mark the construction as elidable
// unconditionally.
if (Entity.allowsNRVO())
CurInit = S.BuildCXXConstructExpr(Loc, Entity.getType(),
Constructor, /*Elidable=*/true,
move_arg(ConstructorArgs),
ConstructorInitRequiresZeroInit,
ConstructKind);
else
CurInit = S.BuildCXXConstructExpr(Loc, Entity.getType(),
Constructor,
move_arg(ConstructorArgs),
ConstructorInitRequiresZeroInit,
ConstructKind);
Add an enum to CXXConstructExpr so we can determine if the construction expression constructs a non-virtual or virtual base. llvm-svn: 102879
2010-05-03 06:54:08 +08:00
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
if (CurInit.isInvalid())
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Access checking for implicit user-defined conversions. llvm-svn: 94971
2010-02-01 11:16:54 +08:00
// Only check access if all of that succeeded.
Pass the InitializedEntity to Sema::CheckConstructorAccess and use it to report different diagnostics depending on which entity is being initialized. llvm-svn: 102010
2010-04-22 02:47:17 +08:00
S.CheckConstructorAccess(Loc, Constructor, Entity,
Remember the "found declaration" for an overload candidate, which is the entity (if applicable) which was actually looked up. If a candidate was found via a using declaration, this is the UsingShadowDecl; otherwise, if the candidate is template specialization, this is the template; otherwise, this is the function. The point of this exercise is that "found declarations" are the entities we do access control for, not their underlying declarations. Broadly speaking, this patch fixes access control for using declarations. There is a *lot* of redundant code calling into the overload-resolution APIs; we really ought to clean that up. llvm-svn: 98945
2010-03-19 15:35:19 +08:00
Step->Function.FoundDecl.getAccess());
Diagnose deprecated/unavailable functions selected by overload resolution. Fixes rdar://problem/4232969, or at least the clang parts of it. llvm-svn: 103191
2010-05-07 02:15:07 +08:00
S.DiagnoseUseOfDecl(Step->Function.FoundDecl, Loc);
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
Rework our handling of copy construction of temporaries, which was a poor (and wrong) approximation of the actual rules governing when to build a copy and when it can be elided. The correct implementation is actually simpler than the approximation. When we only enumerate constructors as part of initialization (e.g., for direct initialization or when we're copying from a class type or one of its derived classes), we don't create a copy. When we enumerate all conversion functions, we do create a copy. Before, we created some extra copies and missed some others. The new test copy-initialization.cpp shows a case where we missed creating a (required, non-elidable) copy as part of a user-defined conversion, which resulted in a miscompile. This commit also fixes PR6757, where the missing copy made us reject well-formed code in the ternary operator. This commit also cleans up our handling of copy elision in the case where we create an extra copy of a temporary object, which became necessary now that we produce the right copies. The code that seeks to find the temporary object being copied has moved into Expr::getTemporaryObject(); it used to have two different not-quite-the-same implementations, one in Sema and one in CodeGen. Note that we still do not attempt to perform the named return value optimization, so we miss copy elisions for return values and throw expressions. llvm-svn: 100196
2010-04-03 02:24:57 +08:00
if (shouldBindAsTemporary(Entity))
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
CurInit = S.MaybeBindToTemporary(CurInit.takeAs<Expr>());
When we create a temporary of class type that we don't immediately bind, check accessibility of the destructor and mark the declaration as referenced. Fixes a bunch of Boost.Regex failures. llvm-svn: 102287
2010-04-25 07:45:46 +08:00
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
break;
}
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
case SK_ZeroInitialization: {
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
step_iterator NextStep = Step;
++NextStep;
if (NextStep != StepEnd &&
NextStep->Kind == SK_ConstructorInitialization) {
// The need for zero-initialization is recorded directly into
// the call to the object's constructor within the next step.
ConstructorInitRequiresZeroInit = true;
} else if (Kind.getKind() == InitializationKind::IK_Value &&
S.getLangOptions().CPlusPlus &&
!Kind.isImplicitValueInit()) {
Provide proper type-source location information for CXXTemporaryObjectExpr, CXXScalarValueInitExpr, and CXXUnresolvedConstructExpr, getting rid of a bunch of FIXMEs in the process. llvm-svn: 113319
2010-09-08 08:15:04 +08:00
TypeSourceInfo *TSInfo = Entity.getTypeSourceInfo();
if (!TSInfo)
TSInfo = S.Context.getTrivialTypeSourceInfo(Step->Type,
Kind.getRange().getBegin());
CurInit = S.Owned(new (S.Context) CXXScalarValueInitExpr(
TSInfo->getType().getNonLValueExprType(S.Context),
TSInfo,
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
Kind.getRange().getEnd()));
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
} else {
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
CurInit = S.Owned(new (S.Context) ImplicitValueInitExpr(Step->Type));
When value-initializing a class with no user-defined constructors but with a non-trivial default constructor, zero-initialize the storage and then call the default constructor. Fixes PR5800. llvm-svn: 91548
2009-12-17 02:50:27 +08:00
}
Implement value initialization in InitializationSequence; untested WIP, yet again. llvm-svn: 91368
2009-12-15 08:01:57 +08:00
break;
}
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
case SK_CAssignment: {
QualType SourceType = CurInitExpr->getType();
Sema::AssignConvertType ConvTy =
S.CheckSingleAssignmentConstraints(Step->Type, CurInitExpr);
Switch initialization of parameters in a call over to InitializationSequence (when a FunctionDecl is present). This required a few small fixes to initialization sequences: - Make sure to use the adjusted parameter type for initialization of function parameters. - Implement transparent union calling semantics in C llvm-svn: 91902
2009-12-22 15:24:36 +08:00
// If this is a call, allow conversion to a transparent union.
if (ConvTy != Sema::Compatible &&
Entity.getKind() == InitializedEntity::EK_Parameter &&
S.CheckTransparentUnionArgumentConstraints(Step->Type, CurInitExpr)
== Sema::Compatible)
ConvTy = Sema::Compatible;
Whenever we complain about a failed initialization of a function or method parameter, provide a note pointing at the parameter itself so the user does not have to manually look for the function/method being called and match up parameters to arguments. For example, we now get: t.c:4:5: warning: incompatible pointer types passing 'long *' to parameter of type 'int *' [-pedantic] f(long_ptr); ^~~~~~~~ t.c:1:13: note: passing argument to parameter 'x' here void f(int *x); ^ llvm-svn: 102038
2010-04-22 08:20:18 +08:00
bool Complained;
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
if (S.DiagnoseAssignmentResult(ConvTy, Kind.getLocation(),
Step->Type, SourceType,
Whenever we complain about a failed initialization of a function or method parameter, provide a note pointing at the parameter itself so the user does not have to manually look for the function/method being called and match up parameters to arguments. For example, we now get: t.c:4:5: warning: incompatible pointer types passing 'long *' to parameter of type 'int *' [-pedantic] f(long_ptr); ^~~~~~~~ t.c:1:13: note: passing argument to parameter 'x' here void f(int *x); ^ llvm-svn: 102038
2010-04-22 08:20:18 +08:00
CurInitExpr,
getAssignmentAction(Entity),
&Complained)) {
PrintInitLocationNote(S, Entity);
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return ExprError();
Whenever we complain about a failed initialization of a function or method parameter, provide a note pointing at the parameter itself so the user does not have to manually look for the function/method being called and match up parameters to arguments. For example, we now get: t.c:4:5: warning: incompatible pointer types passing 'long *' to parameter of type 'int *' [-pedantic] f(long_ptr); ^~~~~~~~ t.c:1:13: note: passing argument to parameter 'x' here void f(int *x); ^ llvm-svn: 102038
2010-04-22 08:20:18 +08:00
} else if (Complained)
PrintInitLocationNote(S, Entity);
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
CurInit.release();
CurInit = S.Owned(CurInitExpr);
break;
}
Initialization improvements: addition of string initialization and a few small bug fixes in SemaInit, switch over SemaDecl to use it more often, and change a bunch of diagnostics which are different with the new initialization code. llvm-svn: 91767
2009-12-19 16:11:05 +08:00
case SK_StringInit: {
QualType Ty = Step->Type;
CheckStringInit(CurInitExpr, ResultType ? *ResultType : Ty, S);
break;
}
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
case SK_ObjCObjectConversion:
S.ImpCastExprToType(CurInitExpr, Step->Type,
GCC didn't care for my attempt at API compatibility, so brute-force everything to the new constants. llvm-svn: 112047
2010-08-25 19:45:40 +08:00
CK_ObjCObjectLValueCast,
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
S.CastCategory(CurInitExpr));
CurInit.release();
CurInit = S.Owned(CurInitExpr);
break;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
}
return move(CurInit);
}
//===----------------------------------------------------------------------===//
// Diagnose initialization failures
//===----------------------------------------------------------------------===//
bool InitializationSequence::Diagnose(Sema &S,
const InitializedEntity &Entity,
const InitializationKind &Kind,
Expr **Args, unsigned NumArgs) {
if (SequenceKind != FailedSequence)
return false;
Switch InitializedEntity from TypeLoc down to just QualTypes, since we don't use the location information but we did spend a bunch of time building faked-up TypeLocs llvm-svn: 91905
2009-12-22 23:35:07 +08:00
QualType DestType = Entity.getType();
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
switch (Failure) {
case FK_TooManyInitsForReference:
Rework base and member initialization in constructors, with several (necessarily simultaneous) changes: - CXXBaseOrMemberInitializer now contains only a single initializer rather than a set of initialiation arguments + a constructor. The single initializer covers all aspects of initialization, including constructor calls as necessary but also cleanup of temporaries created by the initializer (which we never handled before!). - Rework + simplify code generation for CXXBaseOrMemberInitializers, since we can now just emit the initializer as an initializer. - Switched base and member initialization over to the new initialization code (InitializationSequence), so that it - Improved diagnostics for the new initialization code when initializing bases and members, to match the diagnostics produced by the previous (special-purpose) code. - Simplify the representation of type-checked constructor initializers in templates; instead of keeping the fully-type-checked AST, which is rather hard to undo at template instantiation time, throw away the type-checked AST and store the raw expressions in the AST. This simplifies instantiation, but loses a little but of information in the AST. - When type-checking implicit base or member initializers within a dependent context, don't add the generated initializers into the AST, because they'll look like they were explicit. - Record in CXXConstructExpr when the constructor call is to initialize a base class, so that CodeGen does not have to infer it from context. This ensures that we call the right kind of constructor. There are also a few "opportunity" fixes here that were needed to not regress, for example: - Diagnose default-initialization of a const-qualified class that does not have a user-declared default constructor. We had this diagnostic specifically for bases and members, but missed it for variables. That's fixed now. - When defining the implicit constructors, destructor, and copy-assignment operator, set the CurContext to that constructor when we're defining the body. llvm-svn: 94952
2010-01-31 17:12:51 +08:00
// FIXME: Customize for the initialized entity?
if (NumArgs == 0)
S.Diag(Kind.getLocation(), diag::err_reference_without_init)
<< DestType.getNonReferenceType();
else // FIXME: diagnostic below could be better!
S.Diag(Kind.getLocation(), diag::err_reference_has_multiple_inits)
<< SourceRange(Args[0]->getLocStart(), Args[NumArgs - 1]->getLocEnd());
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
case FK_ArrayNeedsInitList:
case FK_ArrayNeedsInitListOrStringLiteral:
S.Diag(Kind.getLocation(), diag::err_array_init_not_init_list)
<< (Failure == FK_ArrayNeedsInitListOrStringLiteral);
break;
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
case FK_AddressOfOverloadFailed: {
DeclAccessPair Found;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
S.ResolveAddressOfOverloadedFunction(Args[0],
DestType.getNonReferenceType(),
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
true,
Found);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
Propagate the "found declaration" (i.e. the using declaration instead of the underlying/instantiated decl) through a lot of API, including "intermediate" MemberExprs required for (e.g.) template instantiation. This is necessary because of the access semantics of member accesses to using declarations: only the base class *containing the using decl* need be accessible from the naming class. This allows us to complete an access-controlled selfhost, if there are no recent regressions. llvm-svn: 99936
2010-03-31 05:47:33 +08:00
}
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
case FK_ReferenceInitOverloadFailed:
Implement user-defined conversions in InitializationSequence. WPI that isn't turned on anyway yet, so it cannot be tested. llvm-svn: 91294
2009-12-15 01:27:33 +08:00
case FK_UserConversionOverloadFailed:
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
switch (FailedOverloadResult) {
case OR_Ambiguous:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
if (Failure == FK_UserConversionOverloadFailed)
S.Diag(Kind.getLocation(), diag::err_typecheck_ambiguous_condition)
<< Args[0]->getType() << DestType
<< Args[0]->getSourceRange();
else
S.Diag(Kind.getLocation(), diag::err_ref_init_ambiguous)
<< DestType << Args[0]->getType()
<< Args[0]->getSourceRange();
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
FailedCandidateSet.NoteCandidates(S, OCD_ViableCandidates, Args, NumArgs);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
case OR_No_Viable_Function:
S.Diag(Kind.getLocation(), diag::err_typecheck_nonviable_condition)
<< Args[0]->getType() << DestType.getNonReferenceType()
<< Args[0]->getSourceRange();
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
FailedCandidateSet.NoteCandidates(S, OCD_AllCandidates, Args, NumArgs);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
case OR_Deleted: {
S.Diag(Kind.getLocation(), diag::err_typecheck_deleted_function)
<< Args[0]->getType() << DestType.getNonReferenceType()
<< Args[0]->getSourceRange();
OverloadCandidateSet::iterator Best;
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
OverloadingResult Ovl
When performing overload resolution, only compare the final conversion sequences for two conversion functions when in fact we are in the text of initialization by a user-defined conversion sequences. Fixes PR8034. llvm-svn: 113724
2010-09-12 16:07:23 +08:00
= FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best,
true);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
if (Ovl == OR_Deleted) {
S.Diag(Best->Function->getLocation(), diag::note_unavailable_here)
<< Best->Function->isDeleted();
} else {
Un-namespace-qualify llvm_unreachable. It's a macro, so the qualification gave no extra safety anyway. llvm-svn: 91207
2009-12-12 13:05:38 +08:00
llvm_unreachable("Inconsistent overload resolution?");
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
break;
}
case OR_Success:
Un-namespace-qualify llvm_unreachable. It's a macro, so the qualification gave no extra safety anyway. llvm-svn: 91207
2009-12-12 13:05:38 +08:00
llvm_unreachable("Conversion did not fail!");
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
break;
}
break;
case FK_NonConstLValueReferenceBindingToTemporary:
case FK_NonConstLValueReferenceBindingToUnrelated:
S.Diag(Kind.getLocation(),
Failure == FK_NonConstLValueReferenceBindingToTemporary
? diag::err_lvalue_reference_bind_to_temporary
: diag::err_lvalue_reference_bind_to_unrelated)
Fix reference-binding when we have a reference to const volatile type; previously, we were allowing this to bind to a temporary. Now, we don't; add test-cases and improve diagnostics. llvm-svn: 94831
2010-01-30 03:39:15 +08:00
<< DestType.getNonReferenceType().isVolatileQualified()
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
<< DestType.getNonReferenceType()
<< Args[0]->getType()
<< Args[0]->getSourceRange();
break;
case FK_RValueReferenceBindingToLValue:
S.Diag(Kind.getLocation(), diag::err_lvalue_to_rvalue_ref)
<< Args[0]->getSourceRange();
break;
case FK_ReferenceInitDropsQualifiers:
S.Diag(Kind.getLocation(), diag::err_reference_bind_drops_quals)
<< DestType.getNonReferenceType()
<< Args[0]->getType()
<< Args[0]->getSourceRange();
break;
case FK_ReferenceInitFailed:
S.Diag(Kind.getLocation(), diag::err_reference_bind_failed)
<< DestType.getNonReferenceType()
<< (Args[0]->isLvalue(S.Context) == Expr::LV_Valid)
<< Args[0]->getType()
<< Args[0]->getSourceRange();
break;
case FK_ConversionFailed:
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
S.Diag(Kind.getLocation(), diag::err_init_conversion_failed)
<< (int)Entity.getKind()
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
<< DestType
<< (Args[0]->isLvalue(S.Context) == Expr::LV_Valid)
<< Args[0]->getType()
<< Args[0]->getSourceRange();
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
break;
case FK_TooManyInitsForScalar: {
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
SourceRange R;
if (InitListExpr *InitList = dyn_cast<InitListExpr>(Args[0]))
Use the new-initialization code for initializing scalars with a function-style cast. Previously, we had a (redundant, incorrect) semantic-checking path for non-class types, which allowed value-initialization of a reference type and then crashed. llvm-svn: 113415
2010-09-09 05:40:08 +08:00
R = SourceRange(InitList->getInit(0)->getLocEnd(),
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
InitList->getLocEnd());
Use the new-initialization code for initializing scalars with a function-style cast. Previously, we had a (redundant, incorrect) semantic-checking path for non-class types, which allowed value-initialization of a reference type and then crashed. llvm-svn: 113415
2010-09-09 05:40:08 +08:00
else
R = SourceRange(Args[0]->getLocEnd(), Args[NumArgs - 1]->getLocEnd());
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
Use the new-initialization code for initializing scalars with a function-style cast. Previously, we had a (redundant, incorrect) semantic-checking path for non-class types, which allowed value-initialization of a reference type and then crashed. llvm-svn: 113415
2010-09-09 05:40:08 +08:00
R.setBegin(S.PP.getLocForEndOfToken(R.getBegin()));
if (Kind.isCStyleOrFunctionalCast())
S.Diag(Kind.getLocation(), diag::err_builtin_func_cast_more_than_one_arg)
<< R;
else
S.Diag(Kind.getLocation(), diag::err_excess_initializers)
<< /*scalar=*/2 << R;
Move initialization via initializer list over to InitializationSequences. llvm-svn: 91050
2009-12-11 01:56:55 +08:00
break;
}
case FK_ReferenceBindingToInitList:
S.Diag(Kind.getLocation(), diag::err_reference_bind_init_list)
<< DestType.getNonReferenceType() << Args[0]->getSourceRange();
break;
case FK_InitListBadDestinationType:
S.Diag(Kind.getLocation(), diag::err_init_list_bad_dest_type)
<< (DestType->isRecordType()) << DestType << Args[0]->getSourceRange();
break;
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
case FK_ConstructorOverloadFailed: {
SourceRange ArgsRange;
if (NumArgs)
ArgsRange = SourceRange(Args[0]->getLocStart(),
Args[NumArgs - 1]->getLocEnd());
// FIXME: Using "DestType" for the entity we're printing is probably
// bad.
switch (FailedOverloadResult) {
case OR_Ambiguous:
S.Diag(Kind.getLocation(), diag::err_ovl_ambiguous_init)
<< DestType << ArgsRange;
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
FailedCandidateSet.NoteCandidates(S, OCD_ViableCandidates,
Args, NumArgs);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
break;
case OR_No_Viable_Function:
Rework base and member initialization in constructors, with several (necessarily simultaneous) changes: - CXXBaseOrMemberInitializer now contains only a single initializer rather than a set of initialiation arguments + a constructor. The single initializer covers all aspects of initialization, including constructor calls as necessary but also cleanup of temporaries created by the initializer (which we never handled before!). - Rework + simplify code generation for CXXBaseOrMemberInitializers, since we can now just emit the initializer as an initializer. - Switched base and member initialization over to the new initialization code (InitializationSequence), so that it - Improved diagnostics for the new initialization code when initializing bases and members, to match the diagnostics produced by the previous (special-purpose) code. - Simplify the representation of type-checked constructor initializers in templates; instead of keeping the fully-type-checked AST, which is rather hard to undo at template instantiation time, throw away the type-checked AST and store the raw expressions in the AST. This simplifies instantiation, but loses a little but of information in the AST. - When type-checking implicit base or member initializers within a dependent context, don't add the generated initializers into the AST, because they'll look like they were explicit. - Record in CXXConstructExpr when the constructor call is to initialize a base class, so that CodeGen does not have to infer it from context. This ensures that we call the right kind of constructor. There are also a few "opportunity" fixes here that were needed to not regress, for example: - Diagnose default-initialization of a const-qualified class that does not have a user-declared default constructor. We had this diagnostic specifically for bases and members, but missed it for variables. That's fixed now. - When defining the implicit constructors, destructor, and copy-assignment operator, set the CurContext to that constructor when we're defining the body. llvm-svn: 94952
2010-01-31 17:12:51 +08:00
if (Kind.getKind() == InitializationKind::IK_Default &&
(Entity.getKind() == InitializedEntity::EK_Base ||
Entity.getKind() == InitializedEntity::EK_Member) &&
isa<CXXConstructorDecl>(S.CurContext)) {
// This is implicit default initialization of a member or
// base within a constructor. If no viable function was
// found, notify the user that she needs to explicitly
// initialize this base/member.
CXXConstructorDecl *Constructor
= cast<CXXConstructorDecl>(S.CurContext);
if (Entity.getKind() == InitializedEntity::EK_Base) {
S.Diag(Kind.getLocation(), diag::err_missing_default_ctor)
<< Constructor->isImplicit()
<< S.Context.getTypeDeclType(Constructor->getParent())
<< /*base=*/0
<< Entity.getType();
RecordDecl *BaseDecl
= Entity.getBaseSpecifier()->getType()->getAs<RecordType>()
->getDecl();
S.Diag(BaseDecl->getLocation(), diag::note_previous_decl)
<< S.Context.getTagDeclType(BaseDecl);
} else {
S.Diag(Kind.getLocation(), diag::err_missing_default_ctor)
<< Constructor->isImplicit()
<< S.Context.getTypeDeclType(Constructor->getParent())
<< /*member=*/1
<< Entity.getName();
S.Diag(Entity.getDecl()->getLocation(), diag::note_field_decl);
if (const RecordType *Record
= Entity.getType()->getAs<RecordType>())
S.Diag(Record->getDecl()->getLocation(),
diag::note_previous_decl)
<< S.Context.getTagDeclType(Record->getDecl());
}
break;
}
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
S.Diag(Kind.getLocation(), diag::err_ovl_no_viable_function_in_init)
<< DestType << ArgsRange;
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
FailedCandidateSet.NoteCandidates(S, OCD_AllCandidates, Args, NumArgs);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
break;
case OR_Deleted: {
S.Diag(Kind.getLocation(), diag::err_ovl_deleted_init)
<< true << DestType << ArgsRange;
OverloadCandidateSet::iterator Best;
Move some of SemaOverload's API to various places in Overload.h, and kill some of it off completely. llvm-svn: 111957
2010-08-25 04:38:10 +08:00
OverloadingResult Ovl
= FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
Constructor initialization for InitializationSequence. Untested WIP. llvm-svn: 91323
2009-12-15 04:49:26 +08:00
if (Ovl == OR_Deleted) {
S.Diag(Best->Function->getLocation(), diag::note_unavailable_here)
<< Best->Function->isDeleted();
} else {
llvm_unreachable("Inconsistent overload resolution?");
}
break;
}
case OR_Success:
llvm_unreachable("Conversion did not fail!");
break;
}
break;
}
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
case FK_DefaultInitOfConst:
Rework base and member initialization in constructors, with several (necessarily simultaneous) changes: - CXXBaseOrMemberInitializer now contains only a single initializer rather than a set of initialiation arguments + a constructor. The single initializer covers all aspects of initialization, including constructor calls as necessary but also cleanup of temporaries created by the initializer (which we never handled before!). - Rework + simplify code generation for CXXBaseOrMemberInitializers, since we can now just emit the initializer as an initializer. - Switched base and member initialization over to the new initialization code (InitializationSequence), so that it - Improved diagnostics for the new initialization code when initializing bases and members, to match the diagnostics produced by the previous (special-purpose) code. - Simplify the representation of type-checked constructor initializers in templates; instead of keeping the fully-type-checked AST, which is rather hard to undo at template instantiation time, throw away the type-checked AST and store the raw expressions in the AST. This simplifies instantiation, but loses a little but of information in the AST. - When type-checking implicit base or member initializers within a dependent context, don't add the generated initializers into the AST, because they'll look like they were explicit. - Record in CXXConstructExpr when the constructor call is to initialize a base class, so that CodeGen does not have to infer it from context. This ensures that we call the right kind of constructor. There are also a few "opportunity" fixes here that were needed to not regress, for example: - Diagnose default-initialization of a const-qualified class that does not have a user-declared default constructor. We had this diagnostic specifically for bases and members, but missed it for variables. That's fixed now. - When defining the implicit constructors, destructor, and copy-assignment operator, set the CurContext to that constructor when we're defining the body. llvm-svn: 94952
2010-01-31 17:12:51 +08:00
if (Entity.getKind() == InitializedEntity::EK_Member &&
isa<CXXConstructorDecl>(S.CurContext)) {
// This is implicit default-initialization of a const member in
// a constructor. Complain that it needs to be explicitly
// initialized.
CXXConstructorDecl *Constructor = cast<CXXConstructorDecl>(S.CurContext);
S.Diag(Kind.getLocation(), diag::err_uninitialized_member_in_ctor)
<< Constructor->isImplicit()
<< S.Context.getTypeDeclType(Constructor->getParent())
<< /*const=*/1
<< Entity.getName();
S.Diag(Entity.getDecl()->getLocation(), diag::note_previous_decl)
<< Entity.getName();
} else {
S.Diag(Kind.getLocation(), diag::err_default_init_const)
<< DestType << (bool)DestType->getAs<RecordType>();
}
Switch the C++ new expression over to InitializationSequence, rather than using its own partial implementation of initialization. Switched CheckInitializerTypes over to InitializedEntity/InitializationKind, to help move us closer to InitializationSequence. Added InitializedEntity::getName() to retrieve the name of the entity, for diagnostics that care about such things. Implemented support for default initialization in InitializationSequence. Clean up the determination of the "source expressions" for an initialization sequence in InitializationSequence::Perform. Taught CXXConstructExpr to store more location information. llvm-svn: 91492
2009-12-16 09:38:02 +08:00
break;
Add a new failure kind, FK_Incomplete, to InitializationSequence, to capture failures when we try to initialize an incomplete type. Previously, we would (ab)use FK_ConversionFailed, then occasionally dereference a null pointer when trying to diagnose the failure. Fixes <rdar://problem/7959007>. llvm-svn: 104286
2010-05-21 06:12:02 +08:00
case FK_Incomplete:
S.RequireCompleteType(Kind.getLocation(), DestType,
diag::err_init_incomplete_type);
break;
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
}
Whenever we complain about a failed initialization of a function or method parameter, provide a note pointing at the parameter itself so the user does not have to manually look for the function/method being called and match up parameters to arguments. For example, we now get: t.c:4:5: warning: incompatible pointer types passing 'long *' to parameter of type 'int *' [-pedantic] f(long_ptr); ^~~~~~~~ t.c:1:13: note: passing argument to parameter 'x' here void f(int *x); ^ llvm-svn: 102038
2010-04-22 08:20:18 +08:00
PrintInitLocationNote(S, Entity);
Reimplement reference initialization (C++ [dcl.init.ref]) using the new notion of an "initialization sequence", which encapsulates the computation of the initialization sequence along with diagnostic information and the capability to turn the computed sequence into an expression. At present, I've only switched one CheckReferenceInit callers over to this new mechanism; more will follow. Aside from (hopefully) being much more true to the standard, the diagnostics provided by this reference-initialization code are a bit better than before. Some examples: p5-var.cpp:54:12: error: non-const lvalue reference to type 'struct Derived' cannot bind to a value of unrelated type 'struct Base' Derived &dr2 = b; // expected-error{{non-const lvalue reference to ... ^ ~ p5-var.cpp:55:9: error: binding of reference to type 'struct Base' to a value of type 'struct Base const' drops qualifiers Base &br3 = bc; // expected-error{{drops qualifiers}} ^ ~~ p5-var.cpp:57:15: error: ambiguous conversion from derived class 'struct Diamond' to base class 'struct Base': struct Diamond -> struct Derived -> struct Base struct Diamond -> struct Derived2 -> struct Base Base &br5 = diamond; // expected-error{{ambiguous conversion from ... ^~~~~~~ p5-var.cpp:59:9: error: non-const lvalue reference to type 'long' cannot bind to a value of unrelated type 'int' long &lr = i; // expected-error{{non-const lvalue reference to type ... ^ ~ p5-var.cpp:74:9: error: non-const lvalue reference to type 'struct Base' cannot bind to a temporary of type 'struct Base' Base &br1 = Base(); // expected-error{{non-const lvalue reference to ... ^ ~~~~~~ p5-var.cpp:102:9: error: non-const reference cannot bind to bit-field 'i' int & ir1 = (ib.i); // expected-error{{non-const reference cannot ... ^ ~~~~~~ p5-var.cpp:98:7: note: bit-field is declared here int i : 17; // expected-note{{bit-field is declared here}} ^ llvm-svn: 90992
2009-12-10 07:02:17 +08:00
return true;
}
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
void InitializationSequence::dump(llvm::raw_ostream &OS) const {
switch (SequenceKind) {
case FailedSequence: {
OS << "Failed sequence: ";
switch (Failure) {
case FK_TooManyInitsForReference:
OS << "too many initializers for reference";
break;
case FK_ArrayNeedsInitList:
OS << "array requires initializer list";
break;
case FK_ArrayNeedsInitListOrStringLiteral:
OS << "array requires initializer list or string literal";
break;
case FK_AddressOfOverloadFailed:
OS << "address of overloaded function failed";
break;
case FK_ReferenceInitOverloadFailed:
OS << "overload resolution for reference initialization failed";
break;
case FK_NonConstLValueReferenceBindingToTemporary:
OS << "non-const lvalue reference bound to temporary";
break;
case FK_NonConstLValueReferenceBindingToUnrelated:
OS << "non-const lvalue reference bound to unrelated type";
break;
case FK_RValueReferenceBindingToLValue:
OS << "rvalue reference bound to an lvalue";
break;
case FK_ReferenceInitDropsQualifiers:
OS << "reference initialization drops qualifiers";
break;
case FK_ReferenceInitFailed:
OS << "reference initialization failed";
break;
case FK_ConversionFailed:
OS << "conversion failed";
break;
case FK_TooManyInitsForScalar:
OS << "too many initializers for scalar";
break;
case FK_ReferenceBindingToInitList:
OS << "referencing binding to initializer list";
break;
case FK_InitListBadDestinationType:
OS << "initializer list for non-aggregate, non-scalar type";
break;
case FK_UserConversionOverloadFailed:
OS << "overloading failed for user-defined conversion";
break;
case FK_ConstructorOverloadFailed:
OS << "constructor overloading failed";
break;
case FK_DefaultInitOfConst:
OS << "default initialization of a const variable";
break;
Add a new failure kind, FK_Incomplete, to InitializationSequence, to capture failures when we try to initialize an incomplete type. Previously, we would (ab)use FK_ConversionFailed, then occasionally dereference a null pointer when trying to diagnose the failure. Fixes <rdar://problem/7959007>. llvm-svn: 104286
2010-05-21 06:12:02 +08:00
case FK_Incomplete:
OS << "initialization of incomplete type";
break;
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
}
OS << '\n';
return;
}
case DependentSequence:
OS << "Dependent sequence: ";
return;
case UserDefinedConversion:
OS << "User-defined conversion sequence: ";
break;
case ConstructorInitialization:
OS << "Constructor initialization sequence: ";
break;
case ReferenceBinding:
OS << "Reference binding: ";
break;
case ListInitialization:
OS << "List initialization: ";
break;
case ZeroInitialization:
OS << "Zero initialization\n";
return;
case NoInitialization:
OS << "No initialization\n";
return;
case StandardConversion:
OS << "Standard conversion: ";
break;
case CAssignment:
OS << "C assignment: ";
break;
case StringInit:
OS << "String initialization: ";
break;
}
for (step_iterator S = step_begin(), SEnd = step_end(); S != SEnd; ++S) {
if (S != step_begin()) {
OS << " -> ";
}
switch (S->Kind) {
case SK_ResolveAddressOfOverloadedFunction:
OS << "resolve address of overloaded function";
break;
case SK_CastDerivedToBaseRValue:
OS << "derived-to-base case (rvalue" << S->Type.getAsString() << ")";
break;
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_CastDerivedToBaseXValue:
OS << "derived-to-base case (xvalue" << S->Type.getAsString() << ")";
break;
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
case SK_CastDerivedToBaseLValue:
OS << "derived-to-base case (lvalue" << S->Type.getAsString() << ")";
break;
case SK_BindReference:
OS << "bind reference to lvalue";
break;
case SK_BindReferenceToTemporary:
OS << "bind reference to a temporary";
break;
In C++98/03, when binding a reference to an rvalue of reference-compatible type, the implementation is permitted to make a copy of the rvalue (or many such copies, even). However, even though we don't make that copy, we are required to check for the presence of a suitable copy constructor. With this change, we do. Note that in C++0x we are not allowed to make these copies, so we test both dialects separately. Also note the FIXME in one of the C++03 tests, where we are not instantiating default function arguments for the copy constructor we pick (but do not call). The fix is obvious; eliminating the infinite recursion it causes is not. Will address that next. llvm-svn: 101704
2010-04-18 15:40:54 +08:00
case SK_ExtraneousCopyToTemporary:
OS << "extraneous C++03 copy to temporary";
break;
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
case SK_UserConversion:
Add raw_ostream operators to NamedDecl for convenience. Switch over all users of getNameAsString on a stream. The next step is to print the name directly into the stream, avoiding a temporary std::string copy. llvm-svn: 101632
2010-04-17 17:33:03 +08:00
OS << "user-defined conversion via " << S->Function.Function;
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
break;
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
case SK_QualificationConversionRValue:
OS << "qualification conversion (rvalue)";
Update ImplicitCastExpr to be able to represent an XValue. llvm-svn: 108807
2010-07-20 12:20:21 +08:00
case SK_QualificationConversionXValue:
OS << "qualification conversion (xvalue)";
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
case SK_QualificationConversionLValue:
OS << "qualification conversion (lvalue)";
break;
case SK_ConversionSequence:
OS << "implicit conversion sequence (";
S->ICS->DebugPrint(); // FIXME: use OS
OS << ")";
break;
case SK_ListInitialization:
OS << "list initialization";
break;
case SK_ConstructorInitialization:
OS << "constructor initialization";
break;
case SK_ZeroInitialization:
OS << "zero initialization";
break;
case SK_CAssignment:
OS << "C assignment";
break;
case SK_StringInit:
OS << "string initialization";
break;
Allow reference binding of a reference of Objective-C object type to an lvalue of another, compatible Objective-C object type (e.g., a subclass). Introduce a new initialization sequence step kind to describe this binding, along with a new cast kind. Fixes PR7741. llvm-svn: 110513
2010-08-07 19:51:51 +08:00
case SK_ObjCObjectConversion:
OS << "Objective-C object conversion";
break;
Fix reference binding of const lvalue references to bit-fields, which requires a temporary. Previously, we were building an initialization sequence that bound to the bit-field as if it were a real lvalue. Note that we previously (and still) diagnose binding of non-const references to bit-fields, as we should. There's no real way to test that this code is correct, since reference binding does not *currently* have any representation in the AST. This fix should make it easier for that to happen, so I've verified this fix with... Added InitializationSequence::dump(), to print an initialization sequence for debugging purposes. llvm-svn: 94826
2010-01-30 03:14:02 +08:00
}
}
}
void InitializationSequence::dump() const {
dump(llvm::errs());
}
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
//===----------------------------------------------------------------------===//
// Initialization helper functions
//===----------------------------------------------------------------------===//
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
Sema::PerformCopyInitialization(const InitializedEntity &Entity,
SourceLocation EqualLoc,
OwningExprResult -> ExprResult. This patch brought to you by M-x query-replace-regexp \(Sema::\|Action::\|Parser::\|\)Owning\(Expr\|Stmt\)Result -> \2Result llvm-svn: 111903
2010-08-24 14:29:42 +08:00
ExprResult Init) {
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
if (Init.isInvalid())
return ExprError();
Expr *InitE = (Expr *)Init.get();
assert(InitE && "No initialization expression?");
if (EqualLoc.isInvalid())
EqualLoc = InitE->getLocStart();
InitializationKind Kind = InitializationKind::CreateCopy(InitE->getLocStart(),
EqualLoc);
InitializationSequence Seq(*this, Entity, Kind, &InitE, 1);
Init.release();
One who seeks knowledge learns something new every day. One who seeks the Tao unlearns something new every day. Less and less remains until you arrive at non-action. When you arrive at non-action, nothing will be left undone. llvm-svn: 112244
2010-08-27 07:41:50 +08:00
return Seq.Perform(*this, Entity, Kind, MultiExprArg(&InitE, 1));
Switch the initialization required by return statements over to the new InitializationSequence. This fixes some bugs (e.g., PR5808), changed some diagnostics, and caused more churn than expected. What's new: - InitializationSequence now has a "C conversion sequence" category and step kind, which falls back to - Changed the diagnostics for returns to always have the result type of the function first and the type of the expression second. CheckSingleAssignmentConstraints to peform checking in C. - Improved ASTs for initialization of return values. The ASTs now capture all of the temporaries we need to create, but intentionally do not bind the tempoary that is actually returned, so that it won't get destroyed twice. - Make sure to perform an (elidable!) copy of the class object that is returned from a class. - Fix copy elision in CodeGen to properly see through the subexpressions that occur with elidable copies. - Give "new" its own entity kind; as with return values and thrown objects, we don't bind the expression so we don't call a destructor for it. Note that, with this patch, I've broken returning move-only types in C++0x. We'll fix it later, when we tackle NRVO. llvm-svn: 91669
2009-12-18 13:02:21 +08:00
}