platform/upstream/llvm.git
11 years ago[msan] Origin stores and loads do not need explicit alignment.
Evgeniy Stepanov [Fri, 14 Dec 2012 13:43:11 +0000 (13:43 +0000)]
[msan] Origin stores and loads do not need explicit alignment.

Origin address is always 4 byte aligned, and the access type is always i32.

llvm-svn: 170199

11 years agoRemove several entries from the 3.2 release notes that were in the 3.1
Chandler Carruth [Fri, 14 Dec 2012 13:41:02 +0000 (13:41 +0000)]
Remove several entries from the 3.2 release notes that were in the 3.1
release notes already. =]

llvm-svn: 170198

11 years agoFix the order of these sections of the release notes.
Chandler Carruth [Fri, 14 Dec 2012 13:37:18 +0000 (13:37 +0000)]
Fix the order of these sections of the release notes.

llvm-svn: 170197

11 years agoAdd a rough draft of some content about the new SROA. I'll try to proof
Chandler Carruth [Fri, 14 Dec 2012 13:37:17 +0000 (13:37 +0000)]
Add a rough draft of some content about the new SROA. I'll try to proof
read this and clean it up tomorrow, but hopefully it's a good
placeholder.

llvm-svn: 170196

11 years agoDelete a long-stale "if its ready" comment. All indications is that this
Chandler Carruth [Fri, 14 Dec 2012 13:37:12 +0000 (13:37 +0000)]
Delete a long-stale "if its ready" comment. All indications is that this
will look a bit different when we have time to get it ready to turn on,
and we won't likely need this reminder.

llvm-svn: 170195

11 years agoProvide some actual highlight bullets for Clang.
Chandler Carruth [Fri, 14 Dec 2012 13:22:57 +0000 (13:22 +0000)]
Provide some actual highlight bullets for Clang.

If anyone has better highlights (I'm obviously biased by the things that
I'm excited about) jump in and add them!

llvm-svn: 170194

11 years ago[asan] more asan_allocator2 code: record the allocation stack trace, implement calloc/etc
Kostya Serebryany [Fri, 14 Dec 2012 13:16:19 +0000 (13:16 +0000)]
[asan] more asan_allocator2 code: record the allocation stack trace, implement calloc/etc

llvm-svn: 170193

11 years ago[msan] Refactor default shadow propagation and origin tracking.
Evgeniy Stepanov [Fri, 14 Dec 2012 12:54:18 +0000 (12:54 +0000)]
[msan] Refactor default shadow propagation and origin tracking.

This change moves the code for default shadow propagaition (handleShadowOr)
and origin tracking (setOriginForNaryOp) into a new builder-like class. Also
gets rid of handleShadowOrBinary.

llvm-svn: 170192

11 years agotsan: fix build
Dmitry Vyukov [Fri, 14 Dec 2012 12:24:11 +0000 (12:24 +0000)]
tsan: fix build

llvm-svn: 170191

11 years ago[asan] more asan_allocator2 code: actually un/poison shadow on malloc/free
Kostya Serebryany [Fri, 14 Dec 2012 12:15:09 +0000 (12:15 +0000)]
[asan] more asan_allocator2 code: actually un/poison shadow on malloc/free

llvm-svn: 170190

11 years ago[Sanitizer] use INTERCEPT_FUNCTION(foo) instead of CHECK(INTERCEPT_FUNCTION(foo)...
Alexey Samsonov [Fri, 14 Dec 2012 11:52:02 +0000 (11:52 +0000)]
[Sanitizer] use INTERCEPT_FUNCTION(foo) instead of CHECK(INTERCEPT_FUNCTION(foo)) - this check may fail on Linux if user program defines its own version of library function foo

llvm-svn: 170189

11 years agoTargets.cpp: [cygwin] Add the predefined macro "_X86_", according to newer version...
NAKAMURA Takumi [Fri, 14 Dec 2012 10:17:26 +0000 (10:17 +0000)]
Targets.cpp: [cygwin] Add the predefined macro "_X86_", according to newer version of cygwin/w32api.

llvm-svn: 170188

11 years agotsan: disable allocator tests in debug build
Dmitry Vyukov [Fri, 14 Dec 2012 10:17:22 +0000 (10:17 +0000)]
tsan: disable allocator tests in debug build

llvm-svn: 170187

11 years agotsan: comment out debug output in test
Dmitry Vyukov [Fri, 14 Dec 2012 10:12:14 +0000 (10:12 +0000)]
tsan: comment out debug output in test

llvm-svn: 170186

11 years agotsan: add sanity test for Go runtime into presubmit test
Dmitry Vyukov [Fri, 14 Dec 2012 10:10:57 +0000 (10:10 +0000)]
tsan: add sanity test for Go runtime into presubmit test

llvm-svn: 170185

11 years agotsan: add socketpair() interceptor and test
Dmitry Vyukov [Fri, 14 Dec 2012 09:57:42 +0000 (09:57 +0000)]
tsan: add socketpair() interceptor and test

llvm-svn: 170184

11 years agoChange TargetLowering::getLoadExtAction to take an MVT, instead of
Patrik Hagglund [Fri, 14 Dec 2012 09:05:13 +0000 (09:05 +0000)]
Change TargetLowering::getLoadExtAction to take an MVT, instead of
EVT.

llvm-svn: 170183

11 years agoHave Sema::ActOnStartOfFunctionDef return the declaration that was passed it.
Argyrios Kyrtzidis [Fri, 14 Dec 2012 06:54:03 +0000 (06:54 +0000)]
Have Sema::ActOnStartOfFunctionDef return the declaration that was passed it.

This fixes the missing warning here:

struct S {
    template <typename T>
    void meth() {
        char arr[3];
        arr[4] = 0; // warning: array index 4 is past the end of the array
    }
};

template <typename T>
void func() {
    char arr[3];
    arr[4] = 0; // no warning
}

llvm-svn: 170180

11 years agoRemove code from Sema::ActOnStartOfFunctionTemplateDef that duplicates what
Argyrios Kyrtzidis [Fri, 14 Dec 2012 06:53:58 +0000 (06:53 +0000)]
Remove code from Sema::ActOnStartOfFunctionTemplateDef that duplicates what
Sema::ActOnStartOfFunctionDef is already doing.

llvm-svn: 170179

11 years agofix another spello
Nico Weber [Fri, 14 Dec 2012 02:41:18 +0000 (02:41 +0000)]
fix another spello

llvm-svn: 170177

11 years agofix spello
Nico Weber [Fri, 14 Dec 2012 02:40:09 +0000 (02:40 +0000)]
fix spello

llvm-svn: 170176

11 years agoclang/test/CodeGenCXX/lambda-expressions.cpp: Relax expression for -Asserts.
NAKAMURA Takumi [Fri, 14 Dec 2012 02:35:04 +0000 (02:35 +0000)]
clang/test/CodeGenCXX/lambda-expressions.cpp: Relax expression for -Asserts.

"entry:" is not met in -Asserts build.

llvm-svn: 170175

11 years agoCleaned up the UUID mismatch just printing itself whenever it wants to by allowing...
Greg Clayton [Fri, 14 Dec 2012 02:15:00 +0000 (02:15 +0000)]
Cleaned up the UUID mismatch just printing itself whenever it wants to by allowing an optional feedback stream to be passed along when getting the symbol vendor.

llvm-svn: 170174

11 years ago<rdar://problem/12878674>
Greg Clayton [Fri, 14 Dec 2012 02:13:44 +0000 (02:13 +0000)]
<rdar://problem/12878674>

Remove debug asserts.

llvm-svn: 170173

11 years agorevert r170166 - disable the loop vectorizer.
Nadav Rotem [Fri, 14 Dec 2012 01:57:00 +0000 (01:57 +0000)]
revert r170166 - disable the loop vectorizer.

llvm-svn: 170172

11 years agoRmoved the old LLVM disassembler based on libedis.
Sean Callanan [Fri, 14 Dec 2012 01:29:59 +0000 (01:29 +0000)]
Rmoved the old LLVM disassembler based on libedis.

llvm-svn: 170171

11 years agoRefactor dump methods to make RegionBindingsRef printable in the debugger.
Ted Kremenek [Fri, 14 Dec 2012 01:23:13 +0000 (01:23 +0000)]
Refactor dump methods to make RegionBindingsRef printable in the debugger.

llvm-svn: 170170

11 years agoMade the struct test case actually use an expression
Sean Callanan [Fri, 14 Dec 2012 00:56:57 +0000 (00:56 +0000)]
Made the struct test case actually use an expression
instead of falling through to the "frame variable"
code.

llvm-svn: 170169

11 years agoSome incorrect debug information caused LLDB
Sean Callanan [Fri, 14 Dec 2012 00:54:13 +0000 (00:54 +0000)]
Some incorrect debug information caused LLDB
to report a structure with an array of size 1
at the end without accounting for that array
when reporting the struct's total size to Clang.

LLDB now coerces such an array to size 0.

<rdar://problem/12822204>

llvm-svn: 170168

11 years agoTrigger the display of error and output in sourced commands from the result object...
Enrico Granata [Fri, 14 Dec 2012 00:52:54 +0000 (00:52 +0000)]
Trigger the display of error and output in sourced commands from the result object's status instead of the presence of text in the error stream
This should be more consistent with the notion of command success/failure and avoids spewing warnings that the user might not care about
There will need to be an option to specify the level of verbosity desired (never show anything, only show failures, errors and warning, everything)

llvm-svn: 170167

11 years agoEnable the loop vectorizer.
Nadav Rotem [Fri, 14 Dec 2012 00:30:34 +0000 (00:30 +0000)]
Enable the loop vectorizer.

llvm-svn: 170166

11 years agoFix a parser_type to get created with the right AST, and also make variables made...
Greg Clayton [Fri, 14 Dec 2012 00:26:21 +0000 (00:26 +0000)]
Fix a parser_type to get created with the right AST, and also make variables made from symbols to not be "void * const", but just "void *".

llvm-svn: 170165

11 years agoMake the test less sensitive to the inline threshold.
Nadav Rotem [Fri, 14 Dec 2012 00:26:15 +0000 (00:26 +0000)]
Make the test less sensitive to the inline threshold.

llvm-svn: 170164

11 years agoFixing the -f option so that one specify multiple filters, e.g.
Enrico Granata [Fri, 14 Dec 2012 00:07:09 +0000 (00:07 +0000)]
Fixing the -f option so that one specify multiple filters, e.g.

./dotest.py  -C clang --arch x86_64 --arch i386  -v -t -f ObjCDataFormatterTestCase.test_appkit_with_dsym_and_run_command -f ObjCDataFormatterTestCase.test_appkit_with_dwarf_and_run_command -f TestObjCBuiltinTypes.test_with_dsym_and_python_api -f TestObjCBuiltinTypes.test_with_dwarf_and_python_api -f ObjCDataFormatterTestCase.test_appkit_with_dsym_and_run_command -f ObjCDataFormatterTestCase.test_appkit_with_dwarf_and_run_command -f TestObjCBuiltinTypes.test_with_dsym_and_python_api -f -TestObjCBuiltinTypes.test_with_dwarf_and_python_api

The previous implementation would only remember the last filter passed, and consequently break redo.py

llvm-svn: 170163

11 years agoDisable the loop vectorizer.
Nadav Rotem [Fri, 14 Dec 2012 00:02:07 +0000 (00:02 +0000)]
Disable the loop vectorizer.

llvm-svn: 170162

11 years ago<rdar://problem/11689939>
Enrico Granata [Thu, 13 Dec 2012 23:50:33 +0000 (23:50 +0000)]
<rdar://problem/11689939>

Supporting a compact display syntax for ObjC pointers where 0x00.....0 is replaced by a much more legible "nil"
e.g. this would show:
(NSArray *) $2 = nil
instead of:
(NSArray *) $2 = 0x0000000000000000 <nil>

llvm-svn: 170161

11 years agoMake sure the __invoke function for lambdas returns properly. Per bug report on...
Eli Friedman [Thu, 13 Dec 2012 23:37:17 +0000 (23:37 +0000)]
Make sure the __invoke function for lambdas returns properly.  Per bug report on IRC>

llvm-svn: 170160

11 years agoUse the new MI bundling API in MachineInstrBundle itself.
Jakob Stoklund Olesen [Thu, 13 Dec 2012 23:23:46 +0000 (23:23 +0000)]
Use the new MI bundling API in MachineInstrBundle itself.

The new API is higher level than just manipulating the bundle flags
directly, and the setIsInsideBundle() function will disappear soon.

llvm-svn: 170159

11 years agoRemove two popcount patterns which we are already able to recognize.
Shuxin Yang [Thu, 13 Dec 2012 23:16:19 +0000 (23:16 +0000)]
Remove two popcount patterns which we are already able to recognize.

llvm-svn: 170158

11 years agoEnable the Loop Vectorizer by default for O2 and O3. Disable if-conversion by default...
Nadav Rotem [Thu, 13 Dec 2012 23:11:54 +0000 (23:11 +0000)]
Enable the Loop Vectorizer by default for O2 and O3. Disable if-conversion by default. I plan to revert this patch later today.

llvm-svn: 170157

11 years agoDebug Info: add support to mark member variables as artificial
David Blaikie [Thu, 13 Dec 2012 22:43:07 +0000 (22:43 +0000)]
Debug Info: add support to mark member variables as artificial

This is the LLVM portion of r170154.

llvm-svn: 170156

11 years agofix comment.
Chris Lattner [Thu, 13 Dec 2012 22:34:43 +0000 (22:34 +0000)]
fix comment.

llvm-svn: 170155

11 years agoDebug Info: Emit vtables pointer members as artificial.
David Blaikie [Thu, 13 Dec 2012 22:29:06 +0000 (22:29 +0000)]
Debug Info: Emit vtables pointer members as artificial.

I wasn't sure where to put the test case for this, but this seemed like as good
a place as any. I had to reorder the tests here to make them legible while
still matching the order of metadata output in the IR file (for some reason
making it virtual changed the ordering).

Relevant commit to fix up LLVM to actually respect 'artificial' member
variables is coming once I write up a test case for it.

llvm-svn: 170154

11 years agoFixed a thinko in the handling of the case where more than one thread had stopped...
Jim Ingham [Thu, 13 Dec 2012 22:24:15 +0000 (22:24 +0000)]
Fixed a thinko in the handling of the case where more than one thread had stopped with real stop reasons at the same time.
Should be that if any of the threads wants to stop, we should stop.  The opposite was what was actually happening

<rdar://problem/12869725>

llvm-svn: 170153

11 years agoRemoved the == and != operators from ArchSpec, since
Sean Callanan [Thu, 13 Dec 2012 22:07:14 +0000 (22:07 +0000)]
Removed the == and != operators from ArchSpec, since
equality can be strict or loose and we want code to
explicitly choose one or the other.

Also renamed the Compare function to IsEqualTo, to
avoid confusion.

<rdar://problem/12856749>

llvm-svn: 170152

11 years ago[PCH] Make the new PCH format (control block) backwards compatible and
Argyrios Kyrtzidis [Thu, 13 Dec 2012 21:38:23 +0000 (21:38 +0000)]
[PCH] Make the new PCH format (control block) backwards compatible and
don't crash when loading a PCH with the older format.

The introduction of the control block broke compatibility with PCHs from
older versions. This patch allows loading (and rejecting) PCHs from an older
version and allows newer PCHs to be rejected from older clang versions as well.

rdar://12821386

llvm-svn: 170150

11 years agoThis is another cleanup patch for 64-bit PowerPC TLS processing. I had
Bill Schmidt [Thu, 13 Dec 2012 20:57:10 +0000 (20:57 +0000)]
This is another cleanup patch for 64-bit PowerPC TLS processing.  I had
some hackery in place that hid my poor use of TblGen, which I've now sorted
out and cleaned up.  No change in observable behavior, so no new test cases.

llvm-svn: 170149

11 years agoChange TargetLowering::setTypeAction to take an MVT, instead fo EVT.
Patrik Hagglund [Thu, 13 Dec 2012 20:42:43 +0000 (20:42 +0000)]
Change TargetLowering::setTypeAction to take an MVT, instead fo EVT.

llvm-svn: 170148

11 years agoAdd support for current Ubuntu Quantal and the upcoming Raring.
Rafael Espindola [Thu, 13 Dec 2012 20:26:05 +0000 (20:26 +0000)]
Add support for current Ubuntu Quantal and the upcoming Raring.
Patch by Martin Nowack.

llvm-svn: 170147

11 years ago<rdar://problem/12700464>
Enrico Granata [Thu, 13 Dec 2012 20:20:11 +0000 (20:20 +0000)]
<rdar://problem/12700464>

Fixing an issue where errors in command files sourced as arguments to command-line lldb (e.g. ./lldb -s foo.cmd) would not be shown to the user

llvm-svn: 170146

11 years agodocs: Improve discussion of syntax highlighting.
Sean Silva [Thu, 13 Dec 2012 20:14:25 +0000 (20:14 +0000)]
docs: Improve discussion of syntax highlighting.

llvm-svn: 170145

11 years agoDocumentation: CompilerWriterInfo.rst: update link to Intel documentation
Dmitri Gribenko [Thu, 13 Dec 2012 20:02:11 +0000 (20:02 +0000)]
Documentation: CompilerWriterInfo.rst: update link to Intel documentation

Replaces old Pentium 4 documentation link with generic current documentation link.

Patch by Kevin Schoedel.

llvm-svn: 170144

11 years agoRevert 170049 because it fails with an assertion on one of the spec2000 workloads.
Nadav Rotem [Thu, 13 Dec 2012 19:58:10 +0000 (19:58 +0000)]
Revert 170049 because it fails with an assertion on one of the spec2000 workloads.

llvm-svn: 170143

11 years agoFix warnings with -DNDEBUG
Tom Stellard [Thu, 13 Dec 2012 19:38:52 +0000 (19:38 +0000)]
Fix warnings with -DNDEBUG

Patch by: NAKAMURA Takumi

llvm-svn: 170142

11 years agoThis is just a clean-up patch that simplifies the initial-exec TLS logic by
Bill Schmidt [Thu, 13 Dec 2012 18:45:54 +0000 (18:45 +0000)]
This is just a clean-up patch that simplifies the initial-exec TLS logic by
avoiding use of machine operand flags.  No change in observable behavior, so
no new test cases.

llvm-svn: 170141

11 years agoChange TargetLowering::getRepRegClassFor to take an MVT, instead of
Patrik Hagglund [Thu, 13 Dec 2012 18:45:35 +0000 (18:45 +0000)]
Change TargetLowering::getRepRegClassFor to take an MVT, instead of
EVT.

Accordingly, change RegDefIter to contain MVTs instead of EVTs.

llvm-svn: 170140

11 years ago[analyzer] Fix doc error (wrong param name) in ObjCSuperCallChecker.
Jordan Rose [Thu, 13 Dec 2012 18:26:05 +0000 (18:26 +0000)]
[analyzer] Fix doc error (wrong param name) in ObjCSuperCallChecker.

Thanks for the -Wdocumentation catch, Dmitri!

llvm-svn: 170139

11 years agoisl: detect vector parallelism
Sebastian Pop [Thu, 13 Dec 2012 16:52:41 +0000 (16:52 +0000)]
isl: detect vector parallelism

llvm-svn: 170138

11 years agoUpdate CMake build corresponding to r170135.
NAKAMURA Takumi [Thu, 13 Dec 2012 16:24:59 +0000 (16:24 +0000)]
Update CMake build corresponding to r170135.

Thanks to Saleem Abdulrasool, aka compnerd!

llvm-svn: 170136

11 years agoDont use/link ARCMT, StaticAnalyzer and Rewriter to clang when the user
Roman Divacky [Thu, 13 Dec 2012 16:09:42 +0000 (16:09 +0000)]
Dont use/link ARCMT, StaticAnalyzer and Rewriter to clang when the user
specifies not to. Dont build ASTMatchers with Rewriter disabled and
StaticAnalyzer when it's disabled.

Without all those three, the clang binary shrinks (x86_64) from ~36MB
to ~32MB (unstripped).

llvm-svn: 170135

11 years agoAdd options to disable building of ARCMT, Rewriter and Static Analyzer
Roman Divacky [Thu, 13 Dec 2012 16:07:19 +0000 (16:07 +0000)]
Add options to disable building of ARCMT, Rewriter and Static Analyzer
in clang. The default remains to build those.

llvm-svn: 170134

11 years agoDocumentation: add AutomaticReferenceCounting.rst to the toctree
Dmitri Gribenko [Thu, 13 Dec 2012 16:06:00 +0000 (16:06 +0000)]
Documentation: add AutomaticReferenceCounting.rst to the toctree

llvm-svn: 170133

11 years agoDocumentation: convert AutomaticReferenceCounting.html to reST
Dmitri Gribenko [Thu, 13 Dec 2012 16:04:37 +0000 (16:04 +0000)]
Documentation: convert AutomaticReferenceCounting.html to reST

Patch by Anastasi Voitova with with small fixes by me.

llvm-svn: 170132

11 years agotsan: fix compilation with -pedantic
Dmitry Vyukov [Thu, 13 Dec 2012 15:26:04 +0000 (15:26 +0000)]
tsan: fix compilation with -pedantic

llvm-svn: 170131

11 years agoFix spelling
Joel Jones [Thu, 13 Dec 2012 15:25:07 +0000 (15:25 +0000)]
Fix spelling

llvm-svn: 170130

11 years agoJITEventListener.h: Use llvm-config.h instead of config.h.
NAKAMURA Takumi [Thu, 13 Dec 2012 15:03:38 +0000 (15:03 +0000)]
JITEventListener.h: Use llvm-config.h instead of config.h.

llvm-svn: 170129

11 years agoRevert r170020, "Simplify negated bit test", for now.
NAKAMURA Takumi [Thu, 13 Dec 2012 14:28:16 +0000 (14:28 +0000)]
Revert r170020, "Simplify negated bit test", for now.

This assumes (1 << n) is always not zero. Consider n is greater than word size.
Although I know it is undefined, this transforms undefined behavior hidden.

This led clang unexpected behavior with some failures. I will investigate to fix undefined shl in clang.

llvm-svn: 170128

11 years agoRename StmtDumper to ASTDumper.
Alexander Kornienko [Thu, 13 Dec 2012 13:59:55 +0000 (13:59 +0000)]
Rename StmtDumper to ASTDumper.
Patch contributed by Philip Craig!

llvm-svn: 170127

11 years agoAdd a dummy documentation file to unbreak 'make install'. We need to find how
Dmitri Gribenko [Thu, 13 Dec 2012 13:37:35 +0000 (13:37 +0000)]
Add a dummy documentation file to unbreak 'make install'.  We need to find how
to package reST-formatted documentation.

llvm-svn: 170126

11 years agoRemove little endian specification from SPIR data layout - SPIR doesn't define endian...
Guy Benyei [Thu, 13 Dec 2012 13:22:48 +0000 (13:22 +0000)]
Remove little endian specification from SPIR data layout - SPIR doesn't define endiannes in the data layout.

llvm-svn: 170125

11 years ago[asan] pop the internal stack frames in SlowUnwindStack, extend the test to check...
Kostya Serebryany [Thu, 13 Dec 2012 12:31:55 +0000 (12:31 +0000)]
[asan] pop the internal stack frames in SlowUnwindStack, extend the test to check this

llvm-svn: 170124

11 years ago[Sanitizer] disable lint check for line length in ASan output tests
Alexey Samsonov [Thu, 13 Dec 2012 12:09:47 +0000 (12:09 +0000)]
[Sanitizer] disable lint check for line length in ASan output tests

llvm-svn: 170123

11 years ago[asan] extend overflow-in-qsort.cc to check both unwind kinds. One more test for...
Kostya Serebryany [Thu, 13 Dec 2012 11:47:49 +0000 (11:47 +0000)]
[asan] extend overflow-in-qsort.cc to check both unwind kinds. One more test for SlowUnwind. Move both to Linux dir (no slow unwind on Mac)

llvm-svn: 170122

11 years ago[ASan] Allow leading underscore in function name to please output tests on Mac
Alexey Samsonov [Thu, 13 Dec 2012 11:11:28 +0000 (11:11 +0000)]
[ASan] Allow leading underscore in function name to please output tests on Mac

llvm-svn: 170121

11 years ago[ASan] don't print memory stats on CheckFailed
Alexey Samsonov [Thu, 13 Dec 2012 11:09:26 +0000 (11:09 +0000)]
[ASan] don't print memory stats on CheckFailed

llvm-svn: 170120

11 years ago[asan] fix win build
Kostya Serebryany [Thu, 13 Dec 2012 10:03:50 +0000 (10:03 +0000)]
[asan] fix win build

llvm-svn: 170119

11 years ago[asan] fix mac build
Kostya Serebryany [Thu, 13 Dec 2012 10:01:20 +0000 (10:01 +0000)]
[asan] fix mac build

llvm-svn: 170118

11 years ago[asan] add two asan flags: fast_unwind_on_fatal and fast_unwind_on_malloc to allow...
Kostya Serebryany [Thu, 13 Dec 2012 09:34:23 +0000 (09:34 +0000)]
[asan] add two asan flags: fast_unwind_on_fatal and fast_unwind_on_malloc to allow using the slow CFI-based unwinder

llvm-svn: 170117

11 years agotsan: completely disable deadlock detector for Go
Dmitry Vyukov [Thu, 13 Dec 2012 09:22:11 +0000 (09:22 +0000)]
tsan: completely disable deadlock detector for Go
seems to cause some weird stack overflow

llvm-svn: 170116

11 years ago[Sanitizer] Add new header - sanitizer_platform_interceptors.h, which should define...
Alexey Samsonov [Thu, 13 Dec 2012 08:50:16 +0000 (08:50 +0000)]
[Sanitizer] Add new header - sanitizer_platform_interceptors.h, which should define whether or not given function should be intercepted on a given platform.

llvm-svn: 170115

11 years ago[ASan] one more macro for 'read' interception
Alexey Samsonov [Thu, 13 Dec 2012 08:36:13 +0000 (08:36 +0000)]
[ASan] one more macro for 'read' interception

llvm-svn: 170114

11 years agotsan: support MapThreadTrace() on all platforms
Dmitry Vyukov [Thu, 13 Dec 2012 08:14:02 +0000 (08:14 +0000)]
tsan: support MapThreadTrace() on all platforms

llvm-svn: 170113

11 years ago[ASan] more macro for conditional interception of pread functions
Alexey Samsonov [Thu, 13 Dec 2012 08:10:23 +0000 (08:10 +0000)]
[ASan] more macro for conditional interception of pread functions

llvm-svn: 170112

11 years ago[asan] add asan/lit_tests/overflow-in-qsort.cc (not fully working yet)
Kostya Serebryany [Thu, 13 Dec 2012 08:05:03 +0000 (08:05 +0000)]
[asan] add asan/lit_tests/overflow-in-qsort.cc (not fully working yet)

llvm-svn: 170111

11 years agoPoint a hyperlink back to where it was before it got reStructured.
Richard Smith [Thu, 13 Dec 2012 07:29:23 +0000 (07:29 +0000)]
Point a hyperlink back to where it was before it got reStructured.

llvm-svn: 170110

11 years agoubsan unit tests for -fsanitize=bool and -fsanitize=enum.
Richard Smith [Thu, 13 Dec 2012 07:12:20 +0000 (07:12 +0000)]
ubsan unit tests for -fsanitize=bool and -fsanitize=enum.

llvm-svn: 170109

11 years agoubsan: Add -fsanitize=bool and -fsanitize=enum, which check for loads of
Richard Smith [Thu, 13 Dec 2012 07:11:50 +0000 (07:11 +0000)]
ubsan: Add -fsanitize=bool and -fsanitize=enum, which check for loads of
bit-patterns which are not valid values for enumerated or boolean types.
These checks are the ubsan analogue of !range metadata.

llvm-svn: 170108

11 years agoubsan: Add -fsanitize=bool and -fsanitize=enum, which check for loads of
Richard Smith [Thu, 13 Dec 2012 07:00:14 +0000 (07:00 +0000)]
ubsan: Add -fsanitize=bool and -fsanitize=enum, which check for loads of
bit-patterns which are not valid values for enumerated or boolean types.
These checks are the ubsan analogue of !range metadata.

llvm-svn: 170107

11 years agoRevert "Restore the PHI optimization I accidently removed" temporarily since
Eric Christopher [Thu, 13 Dec 2012 06:48:05 +0000 (06:48 +0000)]
Revert "Restore the PHI optimization I accidently removed" temporarily since
it seems to be breaking self-host for a few people and is PR14592.

This reverts commit r170024.

llvm-svn: 170106

11 years agoRevert "Add a funciton to get the segment name of a section."
Eric Christopher [Thu, 13 Dec 2012 06:36:18 +0000 (06:36 +0000)]
Revert "Add a funciton to get the segment name of a section."

This reverts commit r170095 since it appears to be breaking the bots.

llvm-svn: 170105

11 years agoChange TargetLowering::getRegClassFor to take an MVT, instead of EVT.
Patrik Hagglund [Thu, 13 Dec 2012 06:34:11 +0000 (06:34 +0000)]
Change TargetLowering::getRegClassFor to take an MVT, instead of EVT.

Accordingly, add helper funtions getSimpleValueType (in parallel to
getValueType) in SDValue, SDNode, and TargetLowering.

This is the first, in a series of patches.

This is the second attempt. In the first attempt (r169837), a few
getSimpleVT() were hoisted too far, detected by bootstrap failures.

llvm-svn: 170104

11 years ago[sanitizer] replace size_t with SIZE_T in interceptors (same for ssize_t, off_t and...
Kostya Serebryany [Thu, 13 Dec 2012 06:31:40 +0000 (06:31 +0000)]
[sanitizer] replace size_t with SIZE_T in interceptors (same for ssize_t, off_t and off64_t). This is done to avoid confusion and to make the code work with compilers that define size_t w/o any includes (MSVC)

llvm-svn: 170103

11 years agoisl: Detect openmp parallelism
Tobias Grosser [Thu, 13 Dec 2012 06:24:06 +0000 (06:24 +0000)]
isl: Detect openmp parallelism

Based on code written by Riyadh Baghdadi.

llvm-svn: 170102

11 years ago[asan] fix windows build
Kostya Serebryany [Thu, 13 Dec 2012 05:51:02 +0000 (05:51 +0000)]
[asan] fix windows build

llvm-svn: 170101

11 years agoSimplify. No functionality change.
Richard Smith [Thu, 13 Dec 2012 05:41:48 +0000 (05:41 +0000)]
Simplify. No functionality change.

llvm-svn: 170100

11 years ago[sanitizer] introduce MmapFixedOrDie and use it in SizeClassAllocator64
Kostya Serebryany [Thu, 13 Dec 2012 05:36:00 +0000 (05:36 +0000)]
[sanitizer] introduce MmapFixedOrDie and use it in SizeClassAllocator64

llvm-svn: 170099

11 years ago[asan] fix the Android build (pread64 interceptor again)
Kostya Serebryany [Thu, 13 Dec 2012 05:27:08 +0000 (05:27 +0000)]
[asan] fix the Android build (pread64 interceptor again)

llvm-svn: 170098

11 years ago[sanitizer] change the way SizeClassAllocator64 allocated memory from the system...
Kostya Serebryany [Thu, 13 Dec 2012 05:05:11 +0000 (05:05 +0000)]
[sanitizer] change the way SizeClassAllocator64 allocated memory from the system: instead of one huge mmap(NORESERVE) it does one huge mprotect and then does small on-demand mmaps. This allows us to call OnMap callbacks which are required to poison newly allocated memory in asan

llvm-svn: 170097

11 years agoInitial support for FreeBSD on ARM.
Rafael Espindola [Thu, 13 Dec 2012 04:17:14 +0000 (04:17 +0000)]
Initial support for FreeBSD on ARM.
Patch by Andrew Turner.

llvm-svn: 170096

11 years agoAdd a funciton to get the segment name of a section.
Rafael Espindola [Thu, 13 Dec 2012 04:07:18 +0000 (04:07 +0000)]
Add a funciton to get the segment name of a section.

On MachO, sections also have segment names. When a tool looking at a .o file
prints a segment name, this is what they mean. In reality, a .o has only one,
anonymous, segment.

This patch adds a MachO only function to fetch that segment name. I named it
getSectionFinalSegmentName since the main use for the name seems to be informing
the linker with segment this section should go to.

The patch also changes MachOObjectFile::getSectionName to return just the
section name instead of computing SegmentName,SectionName.

llvm-svn: 170095