]> granicus.if.org Git - clang/log
clang
14 years agoInclude sender address in completion log.
Ted Kremenek [Thu, 15 Apr 2010 06:32:15 +0000 (06:32 +0000)]
Include sender address in completion log.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101358 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTidy up comment.
Daniel Dunbar [Thu, 15 Apr 2010 06:18:42 +0000 (06:18 +0000)]
Tidy up comment.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101357 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd TargetInfo::useBitfieldTypeAlignment().
Daniel Dunbar [Thu, 15 Apr 2010 06:18:39 +0000 (06:18 +0000)]
Add TargetInfo::useBitfieldTypeAlignment().
 - Used to determine whether the alignment of the type in a bit-field is
   respected when laying out structures. The default is true, targets can
   override this as needed.

 - This is designed to correspond to the PCC_BITFIELD_TYPE_MATTERS macro in
   gcc. The AST/Sema implementation only affects one line, unless I have
   forgotten something. I'd appreciate further review.

 - IRgen still needs to be updated to fully support this (which is effectively
   PR5591).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101356 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDriver/Frontend: Add support for -mllvm, which forwards options to the LLVM option...
Daniel Dunbar [Thu, 15 Apr 2010 06:09:03 +0000 (06:09 +0000)]
Driver/Frontend: Add support for -mllvm, which forwards options to the LLVM option parser.
 - Note that this is a behavior change, previously -mllvm at the driver level forwarded to clang -cc1. The driver does a little magic to make sure that '-mllvm -disable-llvm-optzns' works correctly, but other users will need to be updated to use -Xclang.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101354 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Change CGBitFieldInfo to take the AccessInfo as constructor arguments, it...
Daniel Dunbar [Thu, 15 Apr 2010 05:09:32 +0000 (05:09 +0000)]
IRgen: Change CGBitFieldInfo to take the AccessInfo as constructor arguments, it is now an immutable object.

Also, add some checking of various invariants that should hold on the CGBitFieldInfo access.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101345 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Eliminate now unused fields from CGBitFieldInfo.
Daniel Dunbar [Thu, 15 Apr 2010 05:09:28 +0000 (05:09 +0000)]
IRgen: Eliminate now unused fields from CGBitFieldInfo.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101344 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: (Reapply 101222, with fixes) Move EmitStoreThroughBitfieldLValue to use new...
Daniel Dunbar [Thu, 15 Apr 2010 03:47:33 +0000 (03:47 +0000)]
IRgen: (Reapply 101222, with fixes) Move EmitStoreThroughBitfieldLValue to use new CGBitfieldInfo::AccessInfo decomposition, instead of computing the access policy itself.
 - Sadly, this doesn't seem to give any .ll size win so far. It is possible to make this routine significantly smarter & avoid various shifting, masking, and zext/sext, but I'm not really convinced it is worth it. It is tricky, and this is really instcombine's job.

 - No intended functionality change; the test case is just to increase coverage & serves as a demo file, it worked before this commit.

The new fixes from r101222 are:

 1. The shift to the target position needs to occur after the value is extended to the correct size. This broke Clang bootstrap, among other things no doubt.

 2. Swap the order of arguments to OR, to get a tad more constant folding.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101339 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRewrite handling of 64-bit palignr intrinsics to be vector shuffles.
Eric Christopher [Thu, 15 Apr 2010 01:43:08 +0000 (01:43 +0000)]
Rewrite handling of 64-bit palignr intrinsics to be vector shuffles.
Stop multiplying constant by 8 accordingly in the header and change
intrinsic definition for what types we expect.

Add to existing palignr test to check that we're emitting the correct things.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101332 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTeach ASTVector::append() about the case where 'NumInputs' is 0. This hopefully...
Ted Kremenek [Thu, 15 Apr 2010 01:14:12 +0000 (01:14 +0000)]
Teach ASTVector::append() about the case where 'NumInputs' is 0.  This hopefully fixes
a crash in InitListExpr's ctor.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101328 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd simple python server for recording code completion timings.
Ted Kremenek [Thu, 15 Apr 2010 01:02:31 +0000 (01:02 +0000)]
Add simple python server for recording code completion timings.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101327 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd optional timing logging for code completion results. This causes a UDP packet
Ted Kremenek [Thu, 15 Apr 2010 01:02:28 +0000 (01:02 +0000)]
Add optional timing logging for code completion results.  This causes a UDP packet
containing the time taken for the code completion to be sent to a designated server
(which is specified using a compile-time -D flag).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101326 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoWarn about non-aggregate classes with no user-declared constructors
Douglas Gregor [Thu, 15 Apr 2010 00:00:53 +0000 (00:00 +0000)]
Warn about non-aggregate classes with no user-declared constructors
that have reference or const scalar members, since those members can
never be initializer or modified. Fixes <rdar://problem/7804350>.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101316 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAlways diagnose and complain about problems in
Douglas Gregor [Wed, 14 Apr 2010 23:11:21 +0000 (23:11 +0000)]
Always diagnose and complain about problems in
ResolveAddressOfOverloadedFunction when asked to complain. Previously,
we had some weird handshake where ResolveAddressOfOverloadedFunction
expected its caller to handle some of the diagnostics but not others,
and yet there was no way for the caller to know which case we were
in. Eliminate this madness, fixing <rdar://problem/7765884>.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101312 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix 80-cols violtaions
Sean Hunt [Wed, 14 Apr 2010 23:07:37 +0000 (23:07 +0000)]
Fix 80-cols violtaions

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101311 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoOnce we've emitted a fatal diagnostic, keep counting errors but with a
Douglas Gregor [Wed, 14 Apr 2010 22:19:45 +0000 (22:19 +0000)]
Once we've emitted a fatal diagnostic, keep counting errors but with a
separate count of "suppressed" errors. This way, semantic analysis
bits that depend on the error count to determine whether problems
occured (e.g., some template argument deduction failures, jump-scope
checking) will not get confused.

The actual problem here is that a missing #include (which is a fatal
error) could cause the jump-scope checker to run on invalid code,
which it is not prepared to do. Trivial fix for both
<rdar://problem/7775941> and <rdar://problem/7775709>.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101297 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMake CXLanguage_Invalid the first enum value (with '0' as its value) in CXLanguageKind.
Ted Kremenek [Wed, 14 Apr 2010 20:58:32 +0000 (20:58 +0000)]
Make CXLanguage_Invalid the first enum value (with '0' as its value) in CXLanguageKind.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101287 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix a -pedantic spurious warning involving @dynamic.
Fariborz Jahanian [Wed, 14 Apr 2010 20:52:42 +0000 (20:52 +0000)]
Fix a -pedantic spurious warning involving @dynamic.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101284 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoimprove altivec c++ support by adding casts, patch by
Chris Lattner [Wed, 14 Apr 2010 20:35:39 +0000 (20:35 +0000)]
improve altivec c++ support by adding casts, patch by
Anton Yartsev!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101281 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoThread a Scope pointer into BuildRecoveryCallExpr to help typo
Douglas Gregor [Wed, 14 Apr 2010 20:27:54 +0000 (20:27 +0000)]
Thread a Scope pointer into BuildRecoveryCallExpr to help typo
correction find names when a call failed. Fixes
<rdar://problem/7853795>.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101278 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTeach typo correction about various language keywords. We can't
Douglas Gregor [Wed, 14 Apr 2010 20:04:41 +0000 (20:04 +0000)]
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.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101274 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoReturn the corrected DeclarationName from Sema::CorrectTypo rather
Douglas Gregor [Wed, 14 Apr 2010 17:09:22 +0000 (17:09 +0000)]
Return the corrected DeclarationName from Sema::CorrectTypo rather
than just a bool indicating that correction occurred. No actual
functionality change (it's still always used like a bool), but this
refactoring will be used to support typo correction to keywords.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101259 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen/NeXT: Simplify to use AST record layout for getting offsets instead of the
Daniel Dunbar [Wed, 14 Apr 2010 17:02:21 +0000 (17:02 +0000)]
IRgen/NeXT: Simplify to use AST record layout for getting offsets instead of the
IRgen record layout, which this code doesn't need to depend on.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101257 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoWhen diagnosing suspicious precedence or assignments, move the fix-it
Douglas Gregor [Wed, 14 Apr 2010 16:09:52 +0000 (16:09 +0000)]
When diagnosing suspicious precedence or assignments, move the fix-it
that adds parentheses from the main diagnostic down to a new
note. This way, when the fix-it represents a choice between two
options, each of the options is associted with a note. There is no
default option in such cases. For example:

/Users/dgregor/t.c:2:9: warning: & has lower precedence than ==; ==
will be
      evaluated first [-Wparentheses]
  if (x & y == 0) {
        ^~~~~~~~
/Users/dgregor/t.c:2:9: note: place parentheses around the &
expression to
      evaluate it first
  if (x & y == 0) {
        ^
      (    )
/Users/dgregor/t.c:2:9: note: place parentheses around the ==
expression to
      silence this warning
  if (x & y == 0) {
        ^
          (     )

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101249 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMove GRStmtNodeBuilder::MakeNode() out of line. No functionality change.
Zhongxing Xu [Wed, 14 Apr 2010 06:35:09 +0000 (06:35 +0000)]
Move GRStmtNodeBuilder::MakeNode() out of line. No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101239 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd support for CXXBoolLiteralExpr.
Zhongxing Xu [Wed, 14 Apr 2010 06:29:29 +0000 (06:29 +0000)]
Add support for CXXBoolLiteralExpr.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101238 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUse direct assignment instead of user defined conversion.
Zhongxing Xu [Wed, 14 Apr 2010 05:50:04 +0000 (05:50 +0000)]
Use direct assignment instead of user defined conversion.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101236 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoSpeculatively revert "IRgen: Move EmitStoreThroughBitfieldLValue to use new CGBitfiel...
Daniel Dunbar [Wed, 14 Apr 2010 05:48:35 +0000 (05:48 +0000)]
Speculatively revert "IRgen: Move EmitStoreThroughBitfieldLValue to use new CGBitfieldInfo::AccessInfo decomposition, instead of computing the access policy itself.", I think it might be breaking bootstrap.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101235 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoImprove line marker directive locations, patch by Jordy Rose
Chris Lattner [Wed, 14 Apr 2010 04:28:50 +0000 (04:28 +0000)]
Improve line marker directive locations, patch by Jordy Rose

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101226 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoadd builtin for __builtin_vsprintf, patch by Paul Davey
Chris Lattner [Wed, 14 Apr 2010 04:25:45 +0000 (04:25 +0000)]
add builtin for __builtin_vsprintf, patch by Paul Davey

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101224 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Move EmitStoreThroughBitfieldLValue to use new CGBitfieldInfo::AccessInfo...
Daniel Dunbar [Wed, 14 Apr 2010 04:08:03 +0000 (04:08 +0000)]
IRgen: Move EmitStoreThroughBitfieldLValue to use new CGBitfieldInfo::AccessInfo decomposition, instead of computing the access policy itself.
 - Sadly, this doesn't seem to give any .ll size win so far. It is possible to make this routine significantly smarter & avoid various shifting, masking, and zext/sext, but I'm not really convinced it is worth it. It is tricky, and this is really instcombine's job.

 - No intended functionality change; the test case is just to increase coverage & serves as a demo file, it worked before this commit.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101222 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Tweak CGBitFieldInfo doxyments & add an accessor.
Daniel Dunbar [Wed, 14 Apr 2010 04:07:59 +0000 (04:07 +0000)]
IRgen: Tweak CGBitFieldInfo doxyments & add an accessor.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101221 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agomake the token paste avoidance logic turn "..." into ".. ." instead of ". . ."
Chris Lattner [Wed, 14 Apr 2010 03:57:19 +0000 (03:57 +0000)]
make the token paste avoidance logic turn "..." into ".. ." instead of ". . ."
when avoiding paste.  Patch by David Peixotto!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101218 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoimplement altivec.h and a bunch of support code, patch by Anton Yartsev!
Chris Lattner [Wed, 14 Apr 2010 03:54:58 +0000 (03:54 +0000)]
implement altivec.h and a bunch of support code, patch by Anton Yartsev!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101215 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoremove some obsolete super-handling code that I forgot to zap.
Chris Lattner [Wed, 14 Apr 2010 03:16:52 +0000 (03:16 +0000)]
remove some obsolete super-handling code that I forgot to zap.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101212 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoImplement typo correction for Objective-C message sends when the
Douglas Gregor [Wed, 14 Apr 2010 02:46:37 +0000 (02:46 +0000)]
Implement typo correction for Objective-C message sends when the
receiver is a mis-typed class name. Previously, we would give a non-specific
typo-correction diagnostic from the expression-parsing code, but there
was no fix-it because it was too late to recover. Now, we give a nice
diagnostic

honk.m:6:4: error: unknown receiver 'Hnk'; did you mean 'Honk'?
  [Hnk method];
   ^~~
   Honk
honk.m:1:1: note: 'Honk' declared here
@interface Honk
^

which includes a fix-it.

We still need to recover better from mis-typing "super".

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101211 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIntroduce a parsing action to distinguish between class, instance, and
Douglas Gregor [Wed, 14 Apr 2010 02:22:16 +0000 (02:22 +0000)]
Introduce a parsing action to distinguish between class, instance, and
super message sends in Objective-C. No actual functionality change
here, but it provides a hook so that Sema can typo-correct the
receiver in some cases.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101207 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMark a function declaration invalid if any of its parameter declarations
John McCall [Wed, 14 Apr 2010 01:27:20 +0000 (01:27 +0000)]
Mark a function declaration invalid if any of its parameter declarations
are invalid.  Prevents a crash-on-invalid during template instantiation.
I... really don't understand how this wasn't already present.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101203 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoParse friend template ids as types instead of ending up in
John McCall [Wed, 14 Apr 2010 00:24:33 +0000 (00:24 +0000)]
Parse friend template ids as types instead of ending up in
ActOnClassTemplateSpecialization and being very confused.
Fixes PR6514 (for non-templated-scope friends).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101198 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd encoding of reference types like gcc does for objc methods and
Fariborz Jahanian [Tue, 13 Apr 2010 23:45:47 +0000 (23:45 +0000)]
Add encoding of reference types like gcc does for objc methods and
blocks. Fixes PR6468.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101196 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUse ASTVector instead of std::vector for the Exprs in InitListExpr. Performance
Ted Kremenek [Tue, 13 Apr 2010 23:39:13 +0000 (23:39 +0000)]
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.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101195 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIntroduce ASTVector, which is a std::vector-like class that allocates all memory
Ted Kremenek [Tue, 13 Apr 2010 23:39:09 +0000 (23:39 +0000)]
Introduce ASTVector, which is a std::vector-like class that allocates all memory
using the allocator associated with an ASTContext.  This is largely copy-and-paste
from SmallVector, and should be refactored one day.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101194 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd cursor kind for C++ methods.
Ted Kremenek [Tue, 13 Apr 2010 23:39:06 +0000 (23:39 +0000)]
Add cursor kind for C++ methods.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101193 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Move EmitLoadOfBitfieldLValue to use new CGBitfieldInfo::AccessInfo decomposit...
Daniel Dunbar [Tue, 13 Apr 2010 23:34:15 +0000 (23:34 +0000)]
IRgen: Move EmitLoadOfBitfieldLValue to use new CGBitfieldInfo::AccessInfo decomposition, instead of computing the access policy itself.
 - This lets the method focus slightly more on emitting clean IR to honor the policy which has been selected. On 403.gcc's combine.c, x86_64, -O0, this reduces the number of lines in the .ll file (~= # of instructions) by 2.5%.

 - No intended functionality change -- at -O3 this should produce equivalent if not identical output. On 403.gcc's combine.c, x86_64, -O3, this isn't quite true and some of the changes are regressions, but I'm not going to worry about that until we move to a new access policy.

 - There is still some room for improvement in the generated IR, in particular we can usually fold the sign-extension of the bit-field into one of the component access. See the FIXME.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101192 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix an embarrasing memory error. I was apparently very tired when I wrote this
John McCall [Tue, 13 Apr 2010 22:18:28 +0000 (22:18 +0000)]
Fix an embarrasing memory error.  I was apparently very tired when I wrote this
code the first time.

Fixes PR6827.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101184 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Enhance CGBitFieldInfo with enough information to fully describe the "policy...
Daniel Dunbar [Tue, 13 Apr 2010 20:58:55 +0000 (20:58 +0000)]
IRgen: Enhance CGBitFieldInfo with enough information to fully describe the "policy" with which a bit-field should be accessed.
 - For now, these policies are computed to match the current IRgen strategy, although the new information isn't being used yet (except in -fdump-record-layouts).

 - Design comments appreciated.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101178 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRemove unnecessary cast.
Daniel Dunbar [Tue, 13 Apr 2010 20:52:05 +0000 (20:52 +0000)]
Remove unnecessary cast.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101176 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoSupport befriending members of class template specializations.
John McCall [Tue, 13 Apr 2010 20:37:33 +0000 (20:37 +0000)]
Support befriending members of class template specializations.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101173 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTwik to radar 7839485. Do not attach weak_import attribute to class
Fariborz Jahanian [Tue, 13 Apr 2010 20:22:35 +0000 (20:22 +0000)]
Twik to radar 7839485. Do not attach weak_import attribute to class
for non-fragile abi on darwin.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101168 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTurn off warning about weak_import attribute
Fariborz Jahanian [Tue, 13 Apr 2010 19:02:07 +0000 (19:02 +0000)]
Turn off warning about weak_import attribute
on objc classes for Darwin. Radar 7839485.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101164 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRemoves a FIXME.
Fariborz Jahanian [Tue, 13 Apr 2010 18:43:37 +0000 (18:43 +0000)]
Removes a FIXME.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101161 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoVariation of objc_copyStruct API generation when
Fariborz Jahanian [Tue, 13 Apr 2010 18:32:24 +0000 (18:32 +0000)]
Variation of objc_copyStruct API generation when
property (atomic/nonatomic) is of aggregate type with
gc'able member objects) (NeXT runtime).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101156 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRework the ConstStructBuilder code to emit missing initializer
Chris Lattner [Tue, 13 Apr 2010 18:16:19 +0000 (18:16 +0000)]
Rework the ConstStructBuilder code to emit missing initializer
elements with explicit zero values instead of with tail padding.
On an example like this:

struct foo { int a; int b; };

struct foo fooarray[] = {
    {1, 2},
    {4},
};

We now lay this out as:

@fooarray = global [2 x %struct.foo] [%struct.foo { i32 1, i32 2 }, %struct.foo { i32 4, i32 0 }]

instead of as:

@fooarray = global %0 <{ %struct.foo { i32 1, i32 2 }, %1 { i32 4, [4 x i8] zeroinitializer } }>

Preserving both the struct type of the second element, but also the array type of the entire thing.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101155 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agounbreak tests.
Chris Lattner [Tue, 13 Apr 2010 17:57:11 +0000 (17:57 +0000)]
unbreak tests.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101153 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agomove a bunch of ConstStructBuilder methods out of line.
Chris Lattner [Tue, 13 Apr 2010 17:45:57 +0000 (17:45 +0000)]
move a bunch of ConstStructBuilder methods out of line.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101152 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agominor cleanups
Chris Lattner [Tue, 13 Apr 2010 17:39:09 +0000 (17:39 +0000)]
minor cleanups

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101151 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTeach HasSideEffect about InitListExprs. Not having
Chris Lattner [Tue, 13 Apr 2010 17:34:23 +0000 (17:34 +0000)]
Teach HasSideEffect about InitListExprs.  Not having
this caused us to codegen dead globals like this:

struct foo { int a; int b; };

static struct foo fooarray[] = {
  {1, 2},
  {4},
};

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101150 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agomake the rewriter add a #ifndef around the #define of __attribute__.
Chris Lattner [Tue, 13 Apr 2010 17:33:56 +0000 (17:33 +0000)]
make the rewriter add a #ifndef around the #define of __attribute__.
Without it, there is no reason for a compiler that supports it to
emit the dead static globals that the rewriter labels attribute(used).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101149 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRefactor and simplify the computation of implicit conversion sequences
Douglas Gregor [Tue, 13 Apr 2010 16:31:36 +0000 (16:31 +0000)]
Refactor and simplify the computation of implicit conversion sequences
for reference binding. The code attempted to handle both the
computation of the ICS and the actual conversion, but the latter is an
anachronism: we now use InitializationSequence for that.

Sema::CheckReferenceInit is now a static function TryReferenceInit
that's only use within overload resolution, and has been simplified
slightly. It still needs to be updated per C++ [over.ics.ref], by
eliminating more of the lvalue/rvalue checks.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101136 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoWhen returning the result of a call to an object of class type, do not
Douglas Gregor [Tue, 13 Apr 2010 15:50:39 +0000 (15:50 +0000)]
When returning the result of a call to an object of class type, do not
return a NULL expression; return either an error or a proper
expression. Fixes PR6078.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101133 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDuring referencing binding, only consider conversion functions for
Douglas Gregor [Tue, 13 Apr 2010 15:07:45 +0000 (15:07 +0000)]
During referencing binding, only consider conversion functions for
direct reference binding when the source and target types are not
reference-related. Fixes PR6066.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101132 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMake all cases that we don't handle explicit.
Zhongxing Xu [Tue, 13 Apr 2010 13:15:19 +0000 (13:15 +0000)]
Make all cases that we don't handle explicit.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101129 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDispatch all C++ cast expr to VisitCast().
Zhongxing Xu [Tue, 13 Apr 2010 12:38:32 +0000 (12:38 +0000)]
Dispatch all C++ cast expr to VisitCast().

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101128 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoCFGBuilder: always add C++ member call expr as block-level expr.
Zhongxing Xu [Tue, 13 Apr 2010 09:38:01 +0000 (09:38 +0000)]
CFGBuilder: always add C++ member call expr as block-level expr.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101127 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAllow classes to befriend implicitly-declared members. Fixes PR6207 for
John McCall [Tue, 13 Apr 2010 07:45:41 +0000 (07:45 +0000)]
Allow classes to befriend implicitly-declared members.  Fixes PR6207 for
members of non-templated classes.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101122 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd a cc1 option to specify the max number of nodes the analyzer can explore.
Zhongxing Xu [Tue, 13 Apr 2010 06:44:31 +0000 (06:44 +0000)]
Add a cc1 option to specify the max number of nodes the analyzer can explore.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101120 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoParse constructor names in friend declarations. Part of the fix for
John McCall [Tue, 13 Apr 2010 06:39:49 +0000 (06:39 +0000)]
Parse constructor names in friend declarations.  Part of the fix for
PR6207.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101119 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRemove this test for now until I can figure out
Fariborz Jahanian [Tue, 13 Apr 2010 02:16:35 +0000 (02:16 +0000)]
Remove this test for now until I can figure out
why it keeps failing on buildbot (but not for me).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101115 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDon't try to find a scope corresponding to the search DC for an unfound
John McCall [Tue, 13 Apr 2010 01:44:10 +0000 (01:44 +0000)]
Don't try to find a scope corresponding to the search DC for an unfound
friend declaration;  this used to be important but is now just a waste of time
plus an unreasonable assertion.  Fixes PR6174.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101112 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoModify test to be darwin specific.
Fariborz Jahanian [Tue, 13 Apr 2010 01:10:57 +0000 (01:10 +0000)]
Modify test to be darwin specific.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101109 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoadd frontend support for -fdata-sections and -ffunction-sections,
Chris Lattner [Tue, 13 Apr 2010 00:38:24 +0000 (00:38 +0000)]
add frontend support for -fdata-sections and -ffunction-sections,
patch by Sylvere Teissier!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101108 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd support for objc_copyStruct to enforce
Fariborz Jahanian [Tue, 13 Apr 2010 00:38:05 +0000 (00:38 +0000)]
Add support for objc_copyStruct to enforce
atomicity of aggregate properties in setter/getter
methods. wip.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101107 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix debug info for cleanup block.
Devang Patel [Tue, 13 Apr 2010 00:08:43 +0000 (00:08 +0000)]
Fix debug info for cleanup block.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101100 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agocache the PP's SourceManager.
Chris Lattner [Tue, 13 Apr 2010 00:06:42 +0000 (00:06 +0000)]
cache the PP's SourceManager.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101099 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDiagnose declarations of conversion functions with declarators other than '()'.
John McCall [Tue, 13 Apr 2010 00:04:31 +0000 (00:04 +0000)]
Diagnose declarations of conversion functions with declarators other than '()'.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101098 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agomake the preprocessor listen to linemarker directives in -E mode,
Chris Lattner [Tue, 13 Apr 2010 00:01:41 +0000 (00:01 +0000)]
make the preprocessor listen to linemarker directives in -E mode,
PR6101.  This is based on a patch and testcase by Jordy Rose!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101097 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRemove some dead FIXMEs
Douglas Gregor [Mon, 12 Apr 2010 23:52:24 +0000 (23:52 +0000)]
Remove some dead FIXMEs

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101092 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoImplement C++ [over.ics.user]p3, which restricts the final conversion
Douglas Gregor [Mon, 12 Apr 2010 23:42:09 +0000 (23:42 +0000)]
Implement C++ [over.ics.user]p3, which restricts the final conversion
from a conversion function template specialization to one of exact
match rank. We only know how to test this in C++0x with default
function template arguments, but it's also in the C++03 spec. Fixes
PR6285.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101089 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoImprove source-location information for C++ conversion functions, by
Douglas Gregor [Mon, 12 Apr 2010 23:19:01 +0000 (23:19 +0000)]
Improve source-location information for C++ conversion functions, by
copying the type location information from the conversion-type-id into
the type location information for the function type. Do something
similar for constructors and destructors, by giving their "void"
return type source-location information.

In all of these cases, we previously left this type-source information
uninitialized, which led to various unfortunate crashes.

We still aren't tracking good source-location information for the
actual names. That's PR6357.

John, please check my sanity on this.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101088 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agofix a minor bug I noticed while work with Jordy's patch for PR6101,
Chris Lattner [Mon, 12 Apr 2010 23:04:41 +0000 (23:04 +0000)]
fix a minor bug I noticed while work with Jordy's patch for PR6101,
in an input file like this:

# 42
int x;

we were emitting:

# <something>
 int x;

(with a space before the int) because we weren't clearing the leading
whitespace flag properly after the \n from the directive was handled.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101084 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd fixit hint for missing ':' in ternary expressions.
Ted Kremenek [Mon, 12 Apr 2010 22:10:35 +0000 (22:10 +0000)]
Add fixit hint for missing ':' in ternary expressions.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101073 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agofix PR6814 - Only print [-pedantic] on a diagnostic if -pedantic
Chris Lattner [Mon, 12 Apr 2010 21:53:11 +0000 (21:53 +0000)]
fix PR6814 - Only print [-pedantic] on a diagnostic if -pedantic
actually turned it on.  If a diag is produced by a warning which
is an extension but defaults to on, and has no warning group, don't
print any option info.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101071 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd 'clang_getCursorLanguage' to return the "language" of the AST element (e.g.,...
Ted Kremenek [Mon, 12 Apr 2010 21:22:16 +0000 (21:22 +0000)]
Add 'clang_getCursorLanguage' to return the "language" of the AST element (e.g., distinguish between C and Objective-C language features).  Currently this only returns results for declarations.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101070 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoSort exports file.
Ted Kremenek [Mon, 12 Apr 2010 21:22:10 +0000 (21:22 +0000)]
Sort exports file.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101069 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agofix PR6660/6168: emit padding as zeros instead of undef. Because
Chris Lattner [Mon, 12 Apr 2010 21:10:05 +0000 (21:10 +0000)]
fix PR6660/6168: emit padding as zeros instead of undef.  Because
trailing fields may not be represented in initializer lists, they
are being handled as padding and those fields *must* be zero
initialized.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101067 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Factor out ComputeBitFieldInfo.
Daniel Dunbar [Mon, 12 Apr 2010 21:01:28 +0000 (21:01 +0000)]
IRgen: Factor out ComputeBitFieldInfo.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101066 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoImplement C++ [temp.local]p4, which specifies how we eliminate
Douglas Gregor [Mon, 12 Apr 2010 20:54:26 +0000 (20:54 +0000)]
Implement C++ [temp.local]p4, which specifies how we eliminate
name-lookup ambiguities when there are multiple base classes that are
all specializations of the same class template. This is part of a
general cleanup for ambiguities in template-name lookup. Fixes
PR6717.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101065 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix null dereference in 'WriteSourceLocation' when the FileEntry is null.
Ted Kremenek [Mon, 12 Apr 2010 19:54:17 +0000 (19:54 +0000)]
Fix null dereference in 'WriteSourceLocation' when the FileEntry is null.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101060 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoPrune includes.
Benjamin Kramer [Mon, 12 Apr 2010 19:45:50 +0000 (19:45 +0000)]
Prune includes.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101059 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUpdate doxygen comments about lifetime requirements of CXUnsaveFile data arguments.
Ted Kremenek [Mon, 12 Apr 2010 18:47:26 +0000 (18:47 +0000)]
Update doxygen comments about lifetime requirements of CXUnsaveFile data arguments.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101054 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd some API code for future work.
Fariborz Jahanian [Mon, 12 Apr 2010 18:18:10 +0000 (18:18 +0000)]
Add some API code for future work.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101052 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIRgen: Add CGRecordLayout::dump, and dump (irgen) record layouts as part of -fdump...
Daniel Dunbar [Mon, 12 Apr 2010 18:14:18 +0000 (18:14 +0000)]
IRgen: Add CGRecordLayout::dump, and dump (irgen) record layouts as part of -fdump-record-layouts.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101051 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agofix PR6819
Chris Lattner [Mon, 12 Apr 2010 17:25:51 +0000 (17:25 +0000)]
fix PR6819

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101050 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agotighten the check for cast of super to avoid rejecting valid code,
Chris Lattner [Mon, 12 Apr 2010 17:09:27 +0000 (17:09 +0000)]
tighten the check for cast of super to avoid rejecting valid code,
rdar://7853261

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101048 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoWhen creating the implicitly-declared special member functions, be
Douglas Gregor [Mon, 12 Apr 2010 17:09:20 +0000 (17:09 +0000)]
When creating the implicitly-declared special member functions, be
sure to introduce them into the current Scope (when we have one) in
addition to the DeclContext for the class, so that they can be found
by name lookup for inline members of the class. Fixes PR6570.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101047 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agofix rdar://7852959 - Use of super within a block is actually ok.
Chris Lattner [Mon, 12 Apr 2010 17:03:29 +0000 (17:03 +0000)]
fix rdar://7852959 - Use of super within a block is actually ok.
(aka, Fariborz was right ;-)

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101046 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIssue warning when 'weak_import' attribute is applied on a class only
Fariborz Jahanian [Mon, 12 Apr 2010 16:57:31 +0000 (16:57 +0000)]
Issue warning when 'weak_import' attribute is applied on a class only
when it is not supported in versions of MacOs.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101044 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRemove the blanket statement about Clang not supporting access control. It does now
Douglas Gregor [Mon, 12 Apr 2010 16:01:38 +0000 (16:01 +0000)]
Remove the blanket statement about Clang not supporting access control. It does now

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101041 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix a crash-on-invalid involving name lookup of tag names, where we
Douglas Gregor [Mon, 12 Apr 2010 16:00:01 +0000 (16:00 +0000)]
Fix a crash-on-invalid involving name lookup of tag names, where we
ended up finding a function template that we didn't expect. Recover
more gracefully, and fix a similar issue for class templates.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101040 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd another test case for r101029, which verifies that we now
Douglas Gregor [Mon, 12 Apr 2010 07:51:13 +0000 (07:51 +0000)]
Add another test case for r101029, which verifies that we now
correctly diagnose instantiation of a function parameter with Objective-C
class type (since Objective-C classes can't be passed by value).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101031 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoBe sure to instantiate the parameters of a function, even when the
Douglas Gregor [Mon, 12 Apr 2010 07:48:19 +0000 (07:48 +0000)]
Be sure to instantiate the parameters of a function, even when the
function's type is (strictly speaking) non-dependent. This ensures
that, e.g., default function arguments get instantiated properly.

And, since I couldn't resist, collapse the two implementations of
function-parameter instantiation into calls to a single, new function
(Sema::SubstParmVarDecl), since the two had nearly identical code (and
each had bugs the other didn't!). More importantly, factored out the
semantic analysis of a parameter declaration into
Sema::CheckParameter, which is called both by
Sema::ActOnParamDeclarator (when parameters are parsed) and when a
parameter is instantiated. Previously, we were missing some
Objective-C and address-space checks on instantiated function
parameters.

Fixes PR6733.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@101029 91177308-0d34-0410-b5e6-96231b3b80d8