LHS type and the computation result type; this encodes information into
the AST which is otherwise non-obvious. Fix Sema to always come up with the
right answer for both of these types. Fix IRGen and the analyzer to
account for these changes. This fixes PR2601. The approach is inspired
by PR2601 comment 2.
Note that this changes real *= complex in CodeGen from a silent
miscompilation to an explicit error.
I'm not really sure that the analyzer changes are correct, or how to
test them... someone more familiar with the analyzer should check those
changes.
llvm-svn: 67889
instantiation for C++ typename-specifiers such as
typename T::type
The parsing of typename-specifiers is relatively easy thanks to
annotation tokens. When we see the "typename", we parse the
typename-specifier and produce a typename annotation token. There are
only a few places where we need to handle this. We currently parse the
typename-specifier form that terminates in an identifier, but not the
simple-template-id form, e.g.,
typename T::template apply<U, V>
Parsing of nested-name-specifiers has a similar problem, since at this
point we don't have any representation of a class template
specialization whose template-name is unknown.
Semantic analysis is only partially complete, with some support for
template instantiation that works for simple examples.
llvm-svn: 67875
turned into -m[no-]pascal-strings by the tool chain.
- This still has issue that derived arguments don't propogate "used"
information correctly so spurious "argument unused" warnings will
still show up.
llvm-svn: 67841
- Added an internal helper class 'PathDiagnosticBuilder' which now bundles the
'ExecutionContinues' methods.
- Added preliminary diagnostics for short-circuit '&&' and '||'
llvm-svn: 67822
Anders Johnsen for pointing this out.
- Have PathDiagnosticControlFlowPiece take PathDiagnosticLocation for the
arguments to its constructors.
llvm-svn: 67812
uniqued representation that should both save some memory and make it
far easier to properly build canonical types for types involving
dependent nested-name-specifiers, e.g., "typename T::Nested::type".
This approach will greatly simplify the representation of
CXXScopeSpec. That'll be next.
llvm-svn: 67799
const DiagnosticBuilder &operator<<(const DiagnosticBuilder &DB,
AccessSpecifier AS);
so we can easily add access specifiers to diagnostics.
llvm-svn: 67795
- Added a new class, 'PathDiagnosticLocation', that is a variant for
SourceLocation, SourceRange, or Stmt*. This will be used soon by
PathDiagnosticPieces to describe locations for targets of branches, locations
of events, etc.
- Did some prep. refactoring of PathDiagnosticPieces to prepare them for
adopting the new PathDiagnosticLocation
llvm-svn: 67767
- Temporarily undef'ed __OBJC2__ in nonfragile objc abi mode
as it was forcing ivar synthesis in a certain project which clang
does not yet support.
llvm-svn: 67766
Treat @package the same as @public. The documentation for @package says it is analogous to private_extern for variables/functions. Fully implementing this requires some kind of linker support (so access is denied to code outside the classes executable image). I don't believe GCC fully implements this semantic. Will discuss with Fariborz offline.
llvm-svn: 67755
Zhongxing and I discussed by email.
Main changes:
- Removed SymIntConstraintVal and SymIntConstraint
- Added SymExpr as a parent class to SymbolData, SymSymExpr, SymIntExpr
- Added nonloc::SymExprVal to wrap SymExpr
- SymbolRef is now just a typedef of 'const SymbolData*'
- Bunch of minor code cleanups in how some methods were invoked (no functionality change)
This changes are part of a long-term plan to have full symbolic expression
trees. This will be useful for lazily evaluating complicated expressions.
llvm-svn: 67731
specializations can be treated as a template. Finally, we can parse
and process the first implementation of Fibonacci I wrote!
Note that this code does not handle all of the cases where
injected-class-names can be treated as templates. In particular,
there's an ambiguity case that we should be able to handle (but
can't), e.g.,
template <class T> struct Base { };
template <class T> struct Derived : Base<int>, Base<char> {
typename Derived::Base b; // error: ambiguous
typename Derived::Base<double> d; // OK
};
llvm-svn: 67720
templates, including in-class initializers. For example:
template<typename T, T Divisor>
class X {
public:
static const T value = 10 / Divisor;
};
instantiated with, e.g.,
X<int, 5>::value
to get the value '2'.
llvm-svn: 67715
the declarations of member classes are instantiated when the owning
class template is instantiated. The definitions of such member classes
are instantiated when a complete type is required.
This change also introduces the injected-class-name into a class
template specialization.
llvm-svn: 67707
terminated with an EOF token. The condition it is trying to check for is
handled by this code above.
// Empty arguments are standard in C99 and supported as an extension in
// other modes.
if (ArgTokens.empty() && !Features.C99)
Diag(Tok, diag::ext_empty_fnmacro_arg);
llvm-svn: 67705
- This is really gross, but its the easiest way to match gcc. Once we
are confident in the driver, we can try and push these translations
down into tools.
- No test cases for this yet, it's hard to see the effects of these
translations before the gcc tool argument translation is pulled
over.
- Interaction with "unused argument" warning hasn't been worked out
yet.
- <rdar://problem/6717359> [driver] implement toolchain specific
argument translation.
"It's horrible in here."
llvm-svn: 67683
representing symbolic expressions like 'x'+3 and 'x'+'y'. The design is
subjected to change later when we fix the class hierarchy of symbolic
expressions.
llvm-svn: 67678
- Lift ArgList to a base class for InputArgList and DerivedArgList.
- This is not a great decomposition, but it does embed the
translation into the type system, and keep things efficient for
tool chains that don't want to do any translation.
- No intended functionality change.
Eventually I hope to get rid of tool chain specific translation and
have each tool do the right thing, but for now this is the easiest way
to match gcc precisely (which is good for testing).
llvm-svn: 67676
class C {
C() { }
int a;
};
C::C() : a(10) { }
We also diagnose when initializers are used on declarations that aren't constructors:
t.cpp:1:10: error: only constructors take base initializers
void f() : a(10) { }
^
Doug and/or Sebastian: I'd appreciate a review, especially the nested-name-spec test results (from the looks of it we now match gcc in that test.)
llvm-svn: 67672
failure to perform a declaration. Instead, explicitly note semantic
failures that occur during template parsing with a DeclResult. Fixes
PR3872.
llvm-svn: 67659
of "object type" rather than the C definition of "object type". The
difference is that C's "object type" excludes incomplete types such as
struct X;
However, C's definition also makes it far too easy to use isObjectType
as a means to detect incomplete types when in fact we should use other
means (e.g., Sema::RequireCompleteType) that cope with C++ semantics,
including template instantiation.
I've already audited every use of isObjectType and isIncompleteType to
ensure that they are doing the right thing for both C and C++, so this
is patch does not change any functionality.
llvm-svn: 67648
- -emit-llvm no longer changes what compilation steps are done.
- -emit-llvm and -emit-llvm -S write output files with .o and .s
suffixes, respectively.
- <rdar://problem/6714125> clang-driver should support -O4 and -flto,
like llvm-gcc
llvm-svn: 67645
- Don't default to using clang for C++ (use -ccc-clang-cxx to
override).
- Default to only using clang on i386 and x86_64 (use
-ccc-clang-archs "" to override).
- <rdar://problem/6712350> [driver] clang should not be used on
powerpc by default
- <rdar://problem/6705767> driver should default to -ccc-no-clang-cxx
I plan to add a warning that we are not using the clang compiler for
the given compilation so that users do not think clang is being used
in situations it isn't.
This change is motivated by the desire to be able to drop clang into a
build and have things "just work", even if it happens to get used to
compile C++ code or code for an architecture we don't support yet.
llvm-svn: 67640
- Substitutes both clang and clang-cc.
- Incorporates patch from Jon Simons to diagnose if clang or clang-cc
isn't found.
- Uses full path when running scripts, for more precision in the
output.
llvm-svn: 67610
class C {
void g(C c);
virtual void f() = 0;
};
In this case, C is not known to be abstract when doing semantic analysis on g. This is done by recursively traversing the abstract class and checking the types of member functions.
llvm-svn: 67594
a class template. At present, we can only instantiation normal
methods, but not constructors, destructors, or conversion operators.
As ever, this contains a bit of refactoring in Sema's type-checking. In
particular:
- Split ActOnFunctionDeclarator into ActOnFunctionDeclarator
(handling the declarator itself) and CheckFunctionDeclaration
(checking for the the function declaration), the latter of which
is also used by template instantiation.
- We were performing the adjustment of function parameter types in
three places; collect those into a single new routine.
- When the type of a parameter is adjusted, allocate an
OriginalParmVarDecl to keep track of the type as it was written.
- Eliminate a redundant check for out-of-line declarations of member
functions; hide more C++-specific checks on function declarations
behind if(getLangOptions().CPlusPlus).
llvm-svn: 67575
On a synthetic command line consisting of almost all defined options,
this drops wall time from .00494 to .00336 and user time from .00258
to .00105.
On the same benchmark, clang-driver is about 15% faster than the
primary gcc driver and almost twice as fast as the gcc driver driver.
llvm-svn: 67564
clang doesn't support, and don't want to warn are unused. Eventually
these should disappear.
Here is a more readable list than is in the diff:
W options: -Wall, -Wcast-align, -Wchar-align, -Wchar-subscripts,
-Werror, -Wextra, -Winline, -Wint-to-pointer-cast, -Wmissing-braces,
-Wmost, -Wnested-externs, -Wno-format-y2k, -Wno-four-char-constants,
-Wno-missing-field-initializers, -Wno-trigraphs, -Wno-unknown-pragmas,
-Wno-unused-parameter, -Wparentheses, -Wpointer-arith,
-Wpointer-to-int-cast, -Wreturn-type, -Wshorten-64-to-32, -Wswitch,
-Wunused-function, -Wunused-label, -Wunused-value, -Wunused-variable,
-Wwrite-strings.
f options: -fasm-blocks, -fmessage-length=.
llvm-svn: 67549
library function, accept this declaration and pretend that we do not
know that this is a library function. autoconf depends on this
(broken) behavior.
llvm-svn: 67541
<rdar://problem/6704930> involving SimpleConstraintManager not reasoning well
about symbolic constraint values involving arithmetic operators.
llvm-svn: 67534
incompatibilities in assignments from other pointer incompatibilities.
Based off of the patch in PR3342. (This doesn't implement -Wno-pointer-sign,
but I don't know the driver code very well.)
llvm-svn: 67494
Eli Friedman