Douglas Gregor [Mon, 23 Aug 2010 21:17:50 +0000 (21:17 +0000)]
Introduce a code-completion hook for the Objective-C collection
argument in a for-each statement (e.g., "for (id x in <blah>)"), which
restricts the expression completions provided to Objective-C types (or
class types in C++).
Dan Gohman [Mon, 23 Aug 2010 21:15:56 +0000 (21:15 +0000)]
Reinstate the code for emitting an initial debug type for a struct,
to handle the case where the struct is only forward-declared. In
this case, a temporary MDNode is not needed and not desired.
Tom Care [Mon, 23 Aug 2010 19:51:57 +0000 (19:51 +0000)]
Several small changes to PseudoConstantAnalysis and the way IdempotentOperationChecker uses it.
- Psuedo -> Pseudo (doh...)
- C++ reference support
- Added pseudoconstant test case for __block vars
- Separated out static local checking from pseudoconstant analysis and generalized to non-local checking
- Added missing test cases for storage false positives
Douglas Gregor [Mon, 23 Aug 2010 18:23:48 +0000 (18:23 +0000)]
Introduce a new code-completion point when we're parsing a
declarator. Here, we can only see a few things (e.g., cvr-qualifiers,
nested name specifiers) and we do not want to provide other non-macro
completions. Previously, we would end up in recovery mode and would
provide a large number of non-relevant completions.
Douglas Gregor [Mon, 23 Aug 2010 14:50:27 +0000 (14:50 +0000)]
Emit an error noting that Clang does not support code generation for
the ternary operator without a left-hand side in C++ (PR7726), from
Jean-Daniel Dupas!
Chandler Carruth [Mon, 23 Aug 2010 07:55:51 +0000 (07:55 +0000)]
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.
Chris Lattner [Mon, 23 Aug 2010 05:26:13 +0000 (05:26 +0000)]
fix rdar://8340348, a miscompile of boost that was exposed by r109848.
That revision started classifying truly empty structs like "Y" and "X"
as being NoClass/NoClass and turning them into 'ignore'. The call code
turns around and allocates space for the ignored argument with
GetUndefRValue. The bug is that GetUndefRValue would return the address
as undef, instead of returning an object with a defined address but
undefined contents.
John McCall [Mon, 23 Aug 2010 01:21:21 +0000 (01:21 +0000)]
Abstract out everything having to do with member pointers into the ABI
class; they should just be completely opaque throughout IR gen now,
although I haven't really audited that.
Fix a bug apparently inherited from gcc-4.2 where we failed to null-check
member data pointers when performing derived-to-base or base-to-derived
conversions on them.
John McCall [Sun, 22 Aug 2010 21:01:12 +0000 (21:01 +0000)]
Go back to asking CodeGenTypes whether a type is zero-initializable.
Make CGT defer to the ABI on all member pointer types.
This requires giving CGT a handle to the ABI.
It's way easier to make that work if we avoid lazily creating the ABI.
Make it so.
Douglas Gregor [Sun, 22 Aug 2010 16:15:35 +0000 (16:15 +0000)]
When performing value-initialization for a class with a non-trivial,
implicitly-defined default constructor, zero-initialize the memory
before calling the default constructor. Previously, we would only
zero-initialize in the case of a trivial default constructor.
Also, simplify the hideous logic that determines when we have a
trivial default constructor and, therefore, don't need to emit any
call at all.
John McCall [Sun, 22 Aug 2010 10:59:02 +0000 (10:59 +0000)]
Experiment with using first-class aggregates to represent member function
pointers. I find the resulting code to be substantially cleaner, and it
makes it very easy to use the same APIs for data member pointers (which I have
conscientiously avoided here), and it avoids a plethora of potential
inefficiencies due to excessive memory copying, but we'll have to see if it
actually works.
John McCall [Sat, 21 Aug 2010 22:46:04 +0000 (22:46 +0000)]
The ARM C++ ABI is sufficiently different from the Itanium C++ ABI that
it deserves its own enumerator. Obviously the implementations should
closely follow the Itanium ABI except in cases of divergence.
Zhongxing Xu [Sat, 21 Aug 2010 11:00:26 +0000 (11:00 +0000)]
Remove a special case for OSAtomic functions. We can already bind and retrieve
with the same binding key. The only trick here is that sometimes the Symbolic
region is stored in with an LocAsInteger wrapper. We unwrap that in
SVal::getAsLocSymbol().
Zhongxing Xu [Sat, 21 Aug 2010 06:26:59 +0000 (06:26 +0000)]
When invalidating a struct region, whether its type definition exists is not
relavant any more, because we set its default value to a symbol, and the type of
default symbolic value is irrelavant.
Daniel Dunbar [Sat, 21 Aug 2010 03:15:20 +0000 (03:15 +0000)]
IRgen: Switch more MakeAddr() users to MakeAddrLValue; this time for calls which were previously not computing the qualifier list. In most cases, I don't think it matters, but I believe this is conservatively more correct / consistent.
Daniel Dunbar [Sat, 21 Aug 2010 02:24:36 +0000 (02:24 +0000)]
IRgen: Change Emit{Load,Store}OfScalar to take a required Alignment argument and
update callers as best I can.
- This is a work in progress, our alignment handling is very horrible / sketchy -- I am just aiming for monotonic improvement.
- Serious review appreciated.
Fix an issue with writing to PCH another included PCH, introduced by the "using an AST on-disk hash table for name lookup" commit.
When including a PCH and later re-emitting to another PCH, the name lookup tables of DeclContexts
may be incomplete, since we now lazily deserialize the visible decls of a particular name.
Fix the issue by iterating over the un-deserialized visible decls and completing the lookup tables
of DeclContexts before writing them out.
Dan Gohman [Fri, 20 Aug 2010 22:02:57 +0000 (22:02 +0000)]
Introduce a new temporary MDNode concept. Temporary MDNodes are
not part of the IR, are not uniqued, and may be safely RAUW'd.
This replaces a variety of alternate mechanisms for achieving
the same effect.
objective-c ivar refactoring patch. Iterations
over ivars for a varienty of puposes is now
consolidated into two small routines; DeepCollectObjCIvars
and ShallowCollectObjCIvars.
John McCall [Fri, 20 Aug 2010 18:27:03 +0000 (18:27 +0000)]
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.
Douglas Gregor [Fri, 20 Aug 2010 16:57:37 +0000 (16:57 +0000)]
Fix a major regression with value-initialization of class types with
trivial default constructors. We're weren't zero-initializing them,
which manifested as <rdar://problem/8320532> (a regression in the GCC
test suite) and is likely to have caused significant other breakage.
Chris Lattner [Fri, 20 Aug 2010 16:08:33 +0000 (16:08 +0000)]
fix some vector extractions to return properly zero extended values
(instead of sign extending) to match ICC. GCC is changing this in
a series of their own PRs (e.g. 41323).
Use the AST on-disk hash table for name lookup inside a DeclContext.
*Huge* improvement over the amount of deserializing that we do for C++ lookup.
e.g, if he have the Carbon header precompiled and include it on a file containing this:
Refactoring; move the functionality of ASTWriter::GetOrCreateTypeID to the more generic
MakeTypeID template function which accepts a type and a function object that returns a TypeIdx.
MakeTypeID is in PCHCommon.h so that it can be used by ASTReader too.
projects / clang / log