declaration of that same class that either came from some other module
or occurred in the translation unit loading the module. In this case,
we need to merge the two redeclaration chains immediately so that all
such declarations have the same canonical declaration in the resulting
AST (even though they don't in the module files we've imported).
Focusing on Objective-C classes until I'm happy with the design, then
I'll both (1) extend this notion to other kinds of declarations, and
(2) optimize away this extra checking when we're not dealing with
modules. For now, doing this checking for PCH files/preambles gives us
better testing coverage.
llvm-svn: 147123
redeclaration chains: only ever have the reader search for
redeclarations of the first (canonical) declaration, since we only
ever record redeclaration ranges for the that declaration. Searching
for redeclarations of non-canonical declarations will never find
anything, so it's a complete waste of time.
llvm-svn: 147055
with a definition pointer (e.g., C++ and Objective-C classes), zip
through the redeclaration chain to make sure that all of the
declarations point to the definition data.
As part of this, realized again why the first redeclaration of an
entity in a file is important, and brought back that idea.
llvm-svn: 146886
redeclaration templates (RedeclarableTemplateDecl), similarly to the
way (de-)serialization is implemented for Redeclarable<T>. In the
process, found a simpler formulation for handling redeclaration
chains and implemented that in both places.
The new test establishes that we're building the redeclaration chains
properly. However, the FIXME indicates where we're tickling a
different bug that has to do with us not setting the DefinitionData
pointer properly in redeclarations that we detected after the
definition itself was deserialized. The (separable) fix for that bug
is forthcoming.
llvm-svn: 146883
which there are no redeclarations. This reduced by size of the PCH
file for Cocoa.h by ~650k: ~536k of that was in the new
LOCAL_REDECLARATIONS table, which went from a ridiculous 540k down to
an acceptable 3.5k, while the rest was due to the more compact
abbreviated representation of redeclarable declaration kinds (which no
longer need to store the 'first' declaration ID).
llvm-svn: 146869
variable is initialized by a non-constant expression, and pass in the variable
being declared so that earlier-initialized fields' values can be used.
Rearrange VarDecl init evaluation to make this possible, and in so doing fix a
long-standing issue in our C++ constant expression handling, where we would
mishandle cases like:
extern const int a;
const int n = a;
const int a = 5;
int arr[n];
Here, n is not initialized by a constant expression, so can't be used in an ICE,
even though the initialization expression would be an ICE if it appeared later
in the TU. This requires computing whether the initializer is an ICE eagerly,
and saving that information in PCH files.
llvm-svn: 146856
chains. The previous implementation relied heavily on the declaration
chain being stored as a (circular) linked list on disk, as it is in
memory. However, when deserializing from multiple modules, the
different chains could get mixed up, leading to broken declaration chains.
The new solution keeps track of the first and last declarations in the
chain for each module file. When we load a declaration, we search all
of the module files for redeclarations of that declaration, then
splice together all of the lists into a coherent whole (along with any
redeclarations that were actually parsed).
As a drive-by fix, (de-)serialize the redeclaration chains of
TypedefNameDecls, which had somehow gotten missed previously. Add a
test of this serialization.
This new scheme creates a redeclaration table that is fairly large in
the PCH file (on the order of 400k for Cocoa.h's 12MB PCH file). The
table is mmap'd in and searched via a binary search, but it's still
quite large. A future tweak will eliminate entries for declarations
that have no redeclarations anywhere, and should
drastically reduce the size of this table.
llvm-svn: 146841
applies to an actual definition. Plus, clarify the purpose of this
field and give the accessor a different name, since getLocEnd() is
supposed to be the same as getSourceRange().getEnd().
llvm-svn: 146694
declarations and definitions) as ObjCInterfaceDecls within the same
redeclaration chain. This new representation matches what we do for
C/C++ variables/functions/classes/templates/etc., and makes it
possible to answer the query "where are all of the declarations of
this class?"
llvm-svn: 146679
redeclaration chain for Objective-C classes, including:
- Using the first declaration as the canonical declaration.
- Using the definition as the primary DeclContext
- Making sure that all declarations have a pointer to the definition
data, and the definition knows that it is the definition.
- Serialization support for when a definition gets added to a
declaration that comes from an AST file.
However, note that we're not taking advantage of much of this code
yet, because we're still re-using ObjCInterfaceDecls.
llvm-svn: 146667
separately-allocated DefinitionData structure, which we manage the
same way as CXXRecordDecl::DefinitionData. This prepares the way for
making ObjCInterfaceDecls redeclarable, to more accurately model
forward declarations of Objective-C classes and eliminate the mutation
of ObjCInterfaceDecl that causes us serious trouble in the AST reader.
Note that ObjCInterfaceDecl's accessors are fairly robust against
being applied to forward declarations, because Clang (and Sema in
particular) doesn't perform RequireCompleteType/hasDefinition() checks
everywhere it has to. Each of these overly-robust cases is marked with
a FIXME, which we can tackle over time.
llvm-svn: 146644
library, since modules cut across all of the libraries. Rename
serialization::Module to serialization::ModuleFile to side-step the
annoying naming conflict. Prune a bunch of ModuleMap.h includes that
are no longer needed (most files only needed the Module type).
llvm-svn: 145538
inside an objc container that "contains" other file-level declarations.
When getting the array of file-level declarations that overlap with a file region,
we failed to report that the region overlaps with an objc container, if
the container had other file-level declarations declared lexically inside it.
Fix this by marking such declarations as "isTopLevelDeclInObjCContainer" in the AST
and handling them appropriately.
llvm-svn: 145109
it is going to be rewritten (and the chain will be serialized again), otherwise we may form a cycle in its
categories list when deserializing.
Also introduce ASTMutationListener::CompletedObjCForwardRef to notify that a forward reference
was completed; using Decl's isChangedSinceDeserialization/setChangedSinceDeserialization
is bug inducing and kinda gross, we should phase it out.
Fixes infinite loop in rdar://10418538.
llvm-svn: 144465
of decl bit offsets.
This allows us to easily get at the location of a decl without deserializing it.
It increases size of Cocoa PCH by only 0.2%.
llvm-svn: 143123
essence, the redeclaration chain for a class could end up in an
inconsistent state while deserializing multiple declarations in that
chain, where the circular linked list was not, in fact,
circular. Since only two redeclarations of the same entity will get
loaded when we're in this state, restore circularity when both have
been loaded. Fixes <rdar://problem/10324940> / PR11195.
llvm-svn: 143037
-Add the location of the class name to all objc container decls, not just ObjCInterfaceDecl.
-Make objc decls consistent with the rest of the NamedDecls and have getLocation() point to the
class name, not the location of '@'.
llvm-svn: 141061
Instead of always storing all source locations for the selector identifiers
we check whether all the identifiers are in a "standard" position; "standard" position is
-Immediately before the arguments: -(id)first:(int)x second:(int)y;
-With a space between the arguments: -(id)first: (int)x second: (int)y;
-For nullary selectors, immediately before ';': -(void)release;
In such cases we infer the locations instead of storing them.
llvm-svn: 140989
to the consumer without being fully deserialized).
The regression was on compiling boost.python and it was too difficult to get a reduced
test case unfortunately.
Also modify the logic of how objc methods are getting passed to the consumer;
codegen depended on receiving objc methods before the implementation decl.
Since the interesting objc methods are ones with a body and such methods only
exist inside an ObjCImplDecl, deserialize and pass to consumer all the methods
of ObCImplDecl when we see one.
Fixes http://llvm.org/PR10922 & rdar://10117105.
llvm-svn: 139644
than having CodeGen check whether a declaration comes from an AST file
(which it shouldn't know or care about), make sure that the AST writer and
reader pass along "interesting" declarations that CodeGen needs to
know about.
llvm-svn: 139441
ASTContext reference. Remove all of the extra checking and logic that
was used to cope with a NULL ASTContext. No effective functionality
change.
llvm-svn: 139413
The initial incentive was to fix a crash when PCH chaining categories
to an interface, but the fix was done in the "modules way" that I hear
is popular with the kids these days.
Each module stores the local chain of categories and we combine them
when the interface is loaded. We also warn if non-dependent modules
introduce duplicate named categories.
llvm-svn: 138926
, such as list of forward @class decls, in a DeclGroup
node. Deal with its consequence throught clang. This
is in preparation for more Sema work ahead. // rdar://8843851.
Feel free to reverse if it breaks something important
and I am unavailable.
llvm-svn: 138709
redeclarations of a particular entity would occur in source
order. Friend declarations that occur within class templates (or
member classes thereof) do not follow this, nor would modules. Big
thanks to Erik Verbruggen for reducing this problem from the Very
Large Qt preamble testcase he found.
llvm-svn: 138557
module DAG-based lookup scheme. This required some reshuffling, so
that each module stores its own mapping from DeclContexts to their
lexical and visible sets for those DeclContexts (rather than one big
"chain").
Overall, this allows simple qualified name lookup into the translation
unit to gather results from multiple modules, with the lookup results
in module B shadowing the lookup results in module A when B imports A.
Walking all of the lexical declarations in a module DAG is still a
mess; we'll end up walking the loaded module list backwards, which
works fine for chained PCH but doesn't make sense in a DAG. I'll
tackle this issue as a separate commit.
llvm-svn: 138463
Example:
template <class T>
class A {
public:
template <class U> void f(U p) { }
template <> void f(int p) { } // <== class scope specialization
};
This extension is necessary to parse MSVC standard C++ headers, MFC and ATL code.
BTW, with this feature in, clang can parse (-fsyntax-only) all the MSVC 2010 standard header files without any error.
llvm-svn: 137573
declaration that never actually gets serialized. Instead, serialize
the various kinds of update records (lexical decls, visible decls, the
addition of an anonymous namespace) for the translation unit, even if
we're not chaining. This way, we won't have to deal with multiple
loaded translation unit declarations.
llvm-svn: 137395
IDs will never cross module boundaries, since they're tied to the
CXXDefinitionData, so just use a local mapping throughout. Eliminate
the global -> local tables and supporting data.
llvm-svn: 136847
reader, to allow AST files to be loaded with their declarations
remapped to different ID numbers. Fix a number of places where we were
either failing to map local declaration IDs into global declaration
IDs or where interpreting the local declaration IDs within the wrong
module.
I've tested this via the usual "random gaps" method. It works well
except for the preamble tests, because our handling of the precompiled
preamble requires declaration and preprocessed entity to be stable
when parsing code and then loading that back into memory. This
property will hold in general, but my randomized testing naturally
breaks this property to get more coverage. In the future, I expect
that the precompiled preamble logic won't need this property.
I am very unhappy with the current handling of the translation unit,
which is a rather egregious hack. We're going to have to do something
very different here for loading multiple AST files, because we don't
want to have to cope with merging two translation units. Likely, we'll
just handle translation units entirely via "update" records, and
predefine a single, fixed declaration ID for the translation
unit. That will come later.
llvm-svn: 136779
were (Module*, Offset) with equivalent maps whose value type is just a
Module*. The offsets have moved into corresponding "Base" fields
within the Module itself, where they will also be helpful for
local->global translation (eventually).
llvm-svn: 136441
so that we have one, simple way to map from global bit offsets to
local bit offsets. Eliminates a number of loops over the chain, and
generalizes for more interesting bit remappings.
Also, as an amusing oddity, we were computing global bit offsets
*backwards* for preprocessed entities (e.g., the directly included PCH
file in the chain would start at offset zero, rather than the original
PCH that occurs first in translation unit). Even more amusingly, it
made precompiled preambles work, because we were forgetting to adjust
the local bit offset to a global bit offset when storing preprocessed
entity offsets in the ASTUnit. Two wrongs made a right, and now
they're both right.
llvm-svn: 135750
type IDs into a single place, and make sure that all of the callers
use the appropriate functions to do the mapping. Since the mapping is
still the identity function, this is essentially a no-op.
llvm-svn: 135733
such that every declaration ID loaded from an AST file will go through
a central local -> global mapping function. At present, this change
does nothing, since the local -> global mapping function is the
identity function.
This is the mechanical part of the refactoring; a follow-up patch will
address a few remaining areas where it's not obvious whether we're
dealing with local or global IDs.
llvm-svn: 135711
AST reader down to the AST file + local ID within that file, rather
than lamely walking the PCH chain. There's no actual functionality
change now, but this is cleaner and more general.
llvm-svn: 135548
they should still be officially __strong for the purposes of errors,
block capture, etc. Make a new bit on variables, isARCPseudoStrong(),
and set this for 'self' and these enumeration-loop variables. Change
the code that was looking for the old patterns to look for this bit,
and change IR generation to find this bit and treat the resulting
variable as __unsafe_unretained for the purposes of init/destroy in
the two places it can come up.
llvm-svn: 133243
Related result types apply Cocoa conventions to the type of message
sends and property accesses to Objective-C methods that are known to
always return objects whose type is the same as the type of the
receiving class (or a subclass thereof), such as +alloc and
-init. This tightens up static type safety for Objective-C, so that we
now diagnose mistakes like this:
t.m:4:10: warning: incompatible pointer types initializing 'NSSet *'
with an
expression of type 'NSArray *' [-Wincompatible-pointer-types]
NSSet *array = [[NSArray alloc] init];
^ ~~~~~~~~~~~~~~~~~~~~~~
/System/Library/Frameworks/Foundation.framework/Headers/NSObject.h:72:1:
note:
instance method 'init' is assumed to return an instance of its
receiver
type ('NSArray *')
- (id)init;
^
It also means that we get decent type inference when writing code in
Objective-C++0x:
auto array = [[NSMutableArray alloc] initWithObjects:@"one", @"two",nil];
// ^ now infers NSMutableArray* rather than id
llvm-svn: 132868
hasTrivialDefaultConstructor() really really means it now.
Also implement a fun standards bug regarding aggregates. Doug, if you'd
like, I can un-implement that bug if you think it is truly a defect.
The bug is that non-special-member constructors are never considered
user-provided, so the following is an aggregate:
struct foo {
foo(int);
};
It's kind of bad, but the solution isn't obvious - should
struct foo {
foo (int) = delete;
};
be an aggregate or not?
Lastly, add a missing initialization to FunctionDecl.
llvm-svn: 131101
- New isDefined() function checks for deletedness
- isThisDeclarationADefinition checks for deletedness
- New doesThisDeclarationHaveABody() does what
isThisDeclarationADefinition() used to do
- The IsDeleted bit is not propagated across redeclarations
- isDeleted() now checks the canoncial declaration
- New isDeletedAsWritten() does what it says on the tin.
- isUserProvided() now correct (thanks Richard!)
This fixes the bug that we weren't catching
void foo() = delete;
void foo() {}
as being a redefinition.
llvm-svn: 131013
parameter node and use this to correctly mangle parameter
references in function template signatures.
A follow-up patch will improve the storage usage of these
fields; here I've just done the lazy thing.
llvm-svn: 130669
accompanying fixes to make it work today.
The core of this patch is to provide a link from a TemplateTypeParmType
back to the TemplateTypeParmDecl node which declared it. This in turn
provides much more precise information about the type, where it came
from, and how it functions for AST consumers.
To make the patch work almost a year after its first attempt, it needed
serialization support, and it now retains the old getName() interface.
Finally, it requires us to not attempt to instantiate the type in an
unsupported friend decl -- specifically those coming from template
friend decls but which refer to a specific type through a dependent
name.
A cleaner representation of the last item would be to build
FriendTemplateDecl nodes for these, storing their template parameters
etc, and to perform proper instantation of them like any other template
declaration. They can still be flagged as unsupported for the purpose of
access checking, etc.
This passed an asserts-enabled bootstrap for me, and the reduced test
case mentioned in the original review thread no longer causes issues,
likely fixed at somewhere amidst the 24k revisions that have elapsed.
llvm-svn: 130628
type trait. The previous implementation suffered from several problems:
1) It implemented all of the logic in RecordType by walking over every
base and field in a CXXRecordDecl and validating the constraints of
the standard. This made for very straightforward code, but is
extremely inefficient. It also is conceptually wrong, the logic tied
to the C++ definition of standard-layout classes should be in
CXXRecordDecl, not RecordType.
2) To address the performance problems with #1, a cache bit was added to
CXXRecordDecl, and at the completion of every C++ class, the
RecordType was queried to determine if it was a standard layout
class, and that state was cached. Two things went very very wrong
with this. First, the caching version of the query *was never
called*. Even within the recursive steps of the walk over all fields
and bases the caching variant was not called, making each query
a full *recursive* walk. Second, despite the cache not being used, it
was computed for every class declared, even when the trait was never
used in the program. This probably significantly regressed compile
time performance for edge-case files.
3) An ASTContext was required merely to query the type trait because
querying it performed the actual computations.
4) The caching bit wasn't managed correctly (uninitialized).
The new implementation follows the system for all the other traits on
C++ classes by encoding all the state needed in the definition data and
building up the trait incrementally as each base and member are added to
the definition of the class.
The idiosyncracies of the specification of standard-layout classes
requires more state than I would like; currently 5 bits. I could
eliminate one of the bits easily at the expense of both clarity and
resilience of the code. I might be able to eliminate one of the other
bits by computing its state in terms of other state bits in the
definition. I've already done that in one place where there was a fairly
simple way to achieve it.
It's possible some of the bits could be moved out of the definition data
and into some other structure which isn't serialized if the serialized
bloat is a problem. That would preclude serialization of a partial class
declaration, but that's likely already precluded.
Comments on any of these issues welcome.
llvm-svn: 130601
operators in C++ record declarations.
This patch starts off by updating a bunch of the standard citations to
refer to the draft 0x standard so that the semantics intended for move
varianst is clear. Where necessary these are duplicated so they'll be
available in doxygen.
It adds bit fields to keep track of the state for the move constructs,
and updates all the code necessary to track this state (I think) as
members are declared for a class. It also wires the state into the
various trait-like accessors in the AST's API, and tests that the type
trait expressions now behave correctly in the presence of move
constructors and move assignment operators.
This isn't complete yet due to these glaring FIXMEs:
1) No synthesis of implicit move constructors or assignment operators.
2) I don't think we correctly enforce the new logic for both copy and
move trivial checks: that the *selected* copy/move
constructor/operator is trivial. Currently this requires *all* of them
to be trivial.
3) Some of the trait logic needs to be folded into the fine-grained
trivial bits to more closely match the wording of the standard. For
example, many of the places we currently set a bit to track POD-ness
could be removed by querying other more fine grained traits on
demand.
llvm-svn: 130076
gcc's unused warnings which don't get emitted if the function is referenced even in an unevaluated context
(e.g. in templates, sizeof, etc.). Also, saying that a function is 'unused' because it won't get codegen'ed
is somewhat misleading.
- Don't emit 'unused' warnings for functions that are referenced in any part of the user's code.
- A warning that an internal function/variable won't get emitted is useful though, so introduce
-Wunneeded-internal-declaration which will warn if a function/variable with internal linkage is not
"needed" ('used' from the codegen perspective), e.g:
static void foo() { }
template <int>
void bar() {
foo();
}
test.cpp:1:13: warning: function 'foo' is not needed and will not be emitted
static void foo() { }
^
Addresses rdar://8733476.
llvm-svn: 129794
Douglas Gregor