platform/upstream/llvm.git
11 years ago[mips] Mark pseudo instructions as code-gen only.
Akira Hatanaka [Thu, 8 Aug 2013 21:44:39 +0000 (21:44 +0000)]
[mips] Mark pseudo instructions as code-gen only.

llvm-svn: 188017

11 years ago[mips] Delete register class HWRegs64.
Akira Hatanaka [Thu, 8 Aug 2013 21:37:32 +0000 (21:37 +0000)]
[mips] Delete register class HWRegs64.

No functionality change.

llvm-svn: 188016

11 years agoEnsure that we don't build LLVM with a dependency
Sean Callanan [Thu, 8 Aug 2013 21:30:51 +0000 (21:30 +0000)]
Ensure that we don't build LLVM with a dependency
on curses.

llvm-svn: 188015

11 years agocast fix to appease buildbot
David Fang [Thu, 8 Aug 2013 21:29:30 +0000 (21:29 +0000)]
cast fix to appease buildbot

llvm-svn: 188014

11 years agoObjectiveC migrator: tighten rule about when to
Fariborz Jahanian [Thu, 8 Aug 2013 21:20:01 +0000 (21:20 +0000)]
ObjectiveC migrator: tighten rule about when to
migrate to @property with getter name starting with
'is'.

llvm-svn: 188013

11 years agoRevert "Reapply r185872 now that the address sanitizer has been changed to support...
Arnold Schwaighofer [Thu, 8 Aug 2013 21:04:16 +0000 (21:04 +0000)]
Revert "Reapply r185872 now that the address sanitizer has been changed to support this."

This reverts commit r187939. It broke an O0 build of a spec benchmark.

llvm-svn: 188012

11 years agoUpdate the .drectve section to remove reference to nonexistent files.
Rui Ueyama [Thu, 8 Aug 2013 21:04:09 +0000 (21:04 +0000)]
Update the .drectve section to remove reference to nonexistent files.

llvm-svn: 188011

11 years ago[tests] Remove dead VALGRIND and CLEANED_TESTSUITE makefile variables.
Daniel Dunbar [Thu, 8 Aug 2013 20:59:27 +0000 (20:59 +0000)]
[tests] Remove dead VALGRIND and CLEANED_TESTSUITE makefile variables.

llvm-svn: 188010

11 years ago[lit] Change --show-{tests,suites} to exit after printing.
Daniel Dunbar [Thu, 8 Aug 2013 20:59:25 +0000 (20:59 +0000)]
[lit] Change --show-{tests,suites} to exit after printing.

 - This is a more sensible behavior than printing and also running tests.

llvm-svn: 188009

11 years ago[lit] Remove --repeat option, which wasn't that useful.
Daniel Dunbar [Thu, 8 Aug 2013 20:59:20 +0000 (20:59 +0000)]
[lit] Remove --repeat option, which wasn't that useful.

llvm-svn: 188008

11 years ago[lit] Eliminate some nested imports.
Daniel Dunbar [Thu, 8 Aug 2013 20:59:16 +0000 (20:59 +0000)]
[lit] Eliminate some nested imports.

llvm-svn: 188007

11 years ago[lit] Remove on_clone member, which is no longer used.
Daniel Dunbar [Thu, 8 Aug 2013 20:59:13 +0000 (20:59 +0000)]
[lit] Remove on_clone member, which is no longer used.

llvm-svn: 188006

11 years agoObjectiveC migration: Handle another special case of
Fariborz Jahanian [Thu, 8 Aug 2013 20:51:58 +0000 (20:51 +0000)]
ObjectiveC migration: Handle another special case of
setter/getter methods which can be migrated to
a @property.

llvm-svn: 188005

11 years agoinitial draft of PPCMachObjectWriter.cpp
David Fang [Thu, 8 Aug 2013 20:14:40 +0000 (20:14 +0000)]
initial draft of PPCMachObjectWriter.cpp
this records relocation entries in the mach-o object file
for PIC code generation.
tested on powerpc-darwin8, validated against darwin otool -rvV

llvm-svn: 188004

11 years agoclang-cl: use MS C++ ABI
Hans Wennborg [Thu, 8 Aug 2013 19:54:30 +0000 (19:54 +0000)]
clang-cl: use MS C++ ABI

Differential Revision: http://llvm-reviews.chandlerc.com/D1329

llvm-svn: 188000

11 years agoFix alignof computation of large arrays on x86_64.
Rafael Espindola [Thu, 8 Aug 2013 19:53:46 +0000 (19:53 +0000)]
Fix alignof computation of large arrays on x86_64.

We were exposing the extra alignment given to large arrays. The new behavior
matches gcc, which is a good thing since this is a gcc extension.

Thanks to Joerg Sonnenberger for noticing it.

While at it, centralize the method description in the .h file.

llvm-svn: 187999

11 years agoRevert "clang-cl: Hook up /Za to prevent adding oldnames.lib dependency"
Reid Kleckner [Thu, 8 Aug 2013 19:33:10 +0000 (19:33 +0000)]
Revert "clang-cl: Hook up /Za to prevent adding oldnames.lib dependency"

This reverts commit r187991 and adjusts the comment.  /Za is much more
involved, and we don't want to give anyone the impression we actually
support it.

llvm-svn: 187998

11 years agoMy previous reorganization of addressof broke -std=c++03. Thanks much to Arnold...
Howard Hinnant [Thu, 8 Aug 2013 18:38:55 +0000 (18:38 +0000)]
My previous reorganization of addressof broke -std=c++03.  Thanks much to Arnold Schwaighofer for catching this.  This patch also catches a few more missing addressof in <future>, thanks go to Zhihao Yuan for catching these.

llvm-svn: 187997

11 years agoCheck for division by zero when performing modulus
Sean Callanan [Thu, 8 Aug 2013 17:57:00 +0000 (17:57 +0000)]
Check for division by zero when performing modulus
operations.

<rdar://problem/14656908>

llvm-svn: 187996

11 years agoRemove unreachable code when thread exits on Linux.
Matt Kopec [Thu, 8 Aug 2013 17:55:16 +0000 (17:55 +0000)]
Remove unreachable code when thread exits on Linux.

llvm-svn: 187995

11 years agoAdd clang_Location_isFromMainFile() to libclang.
Stefanus Du Toit [Thu, 8 Aug 2013 17:48:14 +0000 (17:48 +0000)]
Add clang_Location_isFromMainFile() to libclang.

Also bump the minor version number and update libclang.exports.

Reviewed by: Dmitri Gribenko, Doug Gregor

llvm-svn: 187994

11 years agoRemove misleading comment about using cc1 option table.
Hans Wennborg [Thu, 8 Aug 2013 17:34:27 +0000 (17:34 +0000)]
Remove misleading comment about using cc1 option table.

There hasn't been a separate cc1 option table since r155916.

llvm-svn: 187993

11 years agoFix off-by-one error in Regex::isValid
Alexey Samsonov [Thu, 8 Aug 2013 17:32:45 +0000 (17:32 +0000)]
Fix off-by-one error in Regex::isValid

llvm-svn: 187992

11 years agoclang-cl: Hook up /Za to prevent adding oldnames.lib dependency
Hans Wennborg [Thu, 8 Aug 2013 16:54:31 +0000 (16:54 +0000)]
clang-cl: Hook up /Za to prevent adding oldnames.lib dependency

The /Za flag should probably do more, but let's start with this.

Differential Revision: http://llvm-reviews.chandlerc.com/D1320

llvm-svn: 187991

11 years agoFix speling.
David Blaikie [Thu, 8 Aug 2013 16:49:52 +0000 (16:49 +0000)]
Fix speling.

Patch by Richard (legalize@xmission.com)

llvm-svn: 187990

11 years agoscan-build: pass -target through to analyzer
Jordan Rose [Thu, 8 Aug 2013 16:06:26 +0000 (16:06 +0000)]
scan-build: pass -target through to analyzer

llvm-svn: 187989

11 years agoR600/SI: Implement fp32<->fp64 conversions
Niels Ole Salscheider [Thu, 8 Aug 2013 16:06:15 +0000 (16:06 +0000)]
R600/SI: Implement fp32<->fp64 conversions

llvm-svn: 187988

11 years agoR600/SI: Implement sint<->fp64 conversions
Niels Ole Salscheider [Thu, 8 Aug 2013 16:06:08 +0000 (16:06 +0000)]
R600/SI: Implement sint<->fp64 conversions

llvm-svn: 187987

11 years agoUse pop_back() instead of pop_back_val() when the returned value is not used.
Jakub Staszak [Thu, 8 Aug 2013 15:48:46 +0000 (15:48 +0000)]
Use pop_back() instead of pop_back_val() when the returned value is not used.

llvm-svn: 187986

11 years agoRemove the now redundant FeatureFP16 from the Cortex-A15 feature list. It was made...
Silviu Baranga [Thu, 8 Aug 2013 15:47:33 +0000 (15:47 +0000)]
Remove the now redundant FeatureFP16 from the Cortex-A15 feature list. It was made redundant when FeatureVFP4 was added which implies FP16.

llvm-svn: 187985

11 years agoFix the comment.
Jakub Staszak [Thu, 8 Aug 2013 15:19:25 +0000 (15:19 +0000)]
Fix the comment.

llvm-svn: 187984

11 years ago[sanitizer] Limit scandir interceptor to Linux.
Evgeniy Stepanov [Thu, 8 Aug 2013 14:04:36 +0000 (14:04 +0000)]
[sanitizer] Limit scandir interceptor to Linux.

Implementation uses __thread not available on certain platforms.

llvm-svn: 187983

11 years ago[sanitizer] Intercept scandir/scandir64.
Evgeniy Stepanov [Thu, 8 Aug 2013 13:57:15 +0000 (13:57 +0000)]
[sanitizer] Intercept scandir/scandir64.

llvm-svn: 187982

11 years agoPollyDependence: Simplify Read/Write/MayWrite before feeding them into ISL.
Tobias Grosser [Thu, 8 Aug 2013 13:55:45 +0000 (13:55 +0000)]
PollyDependence: Simplify Read/Write/MayWrite before feeding them into ISL.

Contributed-by: Star Tan <tanmx_star@yeah.net>
llvm-svn: 187981

11 years agotsan: better diagnostics for invalid addresses passed to free()
Dmitry Vyukov [Thu, 8 Aug 2013 13:47:50 +0000 (13:47 +0000)]
tsan: better  diagnostics for invalid addresses passed to free()

llvm-svn: 187980

11 years agoIntroduce Replacement deduplication and conflict detection function
Edwin Vane [Thu, 8 Aug 2013 13:31:14 +0000 (13:31 +0000)]
Introduce Replacement deduplication and conflict detection function

Summary:
This patch adds tooling::deduplicate() which removes duplicates from and
looks for conflicts in a vector of Replacements.

Differential Revision: http://llvm-reviews.chandlerc.com/D1314

llvm-svn: 187979

11 years ago[sanitizer] Intercept strerror and strerror_r.
Evgeniy Stepanov [Thu, 8 Aug 2013 11:44:05 +0000 (11:44 +0000)]
[sanitizer] Intercept strerror and strerror_r.

llvm-svn: 187978

11 years agoFactor out parsing sanitizer arguments to a separate function
Alexey Samsonov [Thu, 8 Aug 2013 11:32:17 +0000 (11:32 +0000)]
Factor out parsing sanitizer arguments to a separate function

llvm-svn: 187977

11 years agoUse EXPECT_TRUE/EXPECT_FALSE to avoid bogus warnings from the guts of gtest.
Benjamin Kramer [Thu, 8 Aug 2013 11:17:39 +0000 (11:17 +0000)]
Use EXPECT_TRUE/EXPECT_FALSE to avoid bogus warnings from the guts of gtest.

llvm-svn: 187976

11 years agoRemove unused variable. No functionality change.
Benjamin Kramer [Thu, 8 Aug 2013 11:08:26 +0000 (11:08 +0000)]
Remove unused variable. No functionality change.

llvm-svn: 187975

11 years agotest commit.
Andrea Di Biagio [Thu, 8 Aug 2013 10:46:36 +0000 (10:46 +0000)]
test commit.

llvm-svn: 187974

11 years agoThe name "tCDP" isn't used anywhere else in the source code, so renaming it for consi...
Mihai Popa [Thu, 8 Aug 2013 10:20:41 +0000 (10:20 +0000)]
The name "tCDP" isn't used anywhere else in the source code, so renaming it for consistency doesn't cause any problems.
This is the only Thumb2 instruction defined with "t" prefix; all other Thumb2 instructions have "t2" prefix (e.g. "t2CDP2" which is defined immediately afterwards).

Patch by Artyom Skrobov.

llvm-svn: 187973

11 years agoMove SanitizerArgs implementation from .h to .cpp
Alexey Samsonov [Thu, 8 Aug 2013 10:11:02 +0000 (10:11 +0000)]
Move SanitizerArgs implementation from .h to .cpp

llvm-svn: 187972

11 years ago[sanitizer] ASan/MSan tests for r187967.
Evgeniy Stepanov [Thu, 8 Aug 2013 09:23:19 +0000 (09:23 +0000)]
[sanitizer] ASan/MSan tests for r187967.

r187967: Disable inlining between sanitized and non-sanitized functions.
llvm-svn: 187971

11 years agoDISubprogram metadata have an MDNode representing which base type
Eric Christopher [Thu, 8 Aug 2013 08:35:37 +0000 (08:35 +0000)]
DISubprogram metadata have an MDNode representing which base type
contains the vtable pointer for the class as the operand at index
12. Replace the i32 0 with a null MDNode.

llvm-svn: 187970

11 years agoThe only useful loop unrolling flag to give realistically is
Chandler Carruth [Thu, 8 Aug 2013 08:34:35 +0000 (08:34 +0000)]
The only useful loop unrolling flag to give realistically is
'-fno-unroll-loops'. The option to the backend is even called
'DisableUnrollLoops'. This is precisely the form that Clang *didn't*
support. We didn't recognize the flag, we didn't pass it to the CC1
layer, and even if we did we wouldn't use it. Clang only inspected the
positive form of the flag, and only did so to enable loop unrolling when
the optimization level wasn't high enough. This only occurs for an
optimization level that even has a chance of running the loop unroller
when optimizing for size.

This commit wires up the 'no' variant, and switches the code to actually
follow the standard flag pattern of using the last flag and allowing
a flag in either direction to override the default.

I think this is still wrong. I don't know why we disable the loop
unroller entirely *from Clang* when optimizing for size, as the loop
unrolling pass *already has special logic* for the case where the
function is attributed as optimized for size! We should really be
trusting that. Maybe in a follow-up patch, I don't really want to change
behavior here.

llvm-svn: 187969

11 years agoSimplify (and fix) check-dfsan configs
Alexey Samsonov [Thu, 8 Aug 2013 08:26:39 +0000 (08:26 +0000)]
Simplify (and fix) check-dfsan configs

llvm-svn: 187968

11 years agoDisable inlining between sanitized and non-sanitized functions.
Evgeniy Stepanov [Thu, 8 Aug 2013 08:22:39 +0000 (08:22 +0000)]
Disable inlining between sanitized and non-sanitized functions.

Inlining between functions with different values of sanitize_* attributes
leads to over- or under-sanitizing, which is always bad.

llvm-svn: 187967

11 years agoFor DW_TAG_template_type_parameter the actual passed in type could
Eric Christopher [Thu, 8 Aug 2013 08:09:43 +0000 (08:09 +0000)]
For DW_TAG_template_type_parameter the actual passed in type could
be void and therefore not have a type entry. Only add the type if
it is non-void and provide a testcase.

llvm-svn: 187966

11 years agoRemove AllUndef check from one of the loops in getVectorShuffle. It was already handl...
Craig Topper [Thu, 8 Aug 2013 08:03:12 +0000 (08:03 +0000)]
Remove AllUndef check from one of the loops in getVectorShuffle. It was already handled by the 'AllLHS && AllRHS' check after the previous loop.

llvm-svn: 187965

11 years agoThe conversion to bool is fine here, no need to check isType.
Eric Christopher [Thu, 8 Aug 2013 07:40:42 +0000 (07:40 +0000)]
The conversion to bool is fine here, no need to check isType.

llvm-svn: 187964

11 years agoMake sure that if we're going to attempt to add a type to a DIE that
Eric Christopher [Thu, 8 Aug 2013 07:40:37 +0000 (07:40 +0000)]
Make sure that if we're going to attempt to add a type to a DIE that
the type exists.

Fix up cases where we weren't checking for optional types and add
an assert to addType to make sure we catch this in the future.

Fix up a testcase that was using the tag for DW_TAG_array_type
when it meant DW_TAG_enumeration_type.

llvm-svn: 187963

11 years agoChange variable name and reflow formatting.
Eric Christopher [Thu, 8 Aug 2013 07:40:31 +0000 (07:40 +0000)]
Change variable name and reflow formatting.

llvm-svn: 187962

11 years agoOptimize mask generation for one of the DAG combiner shufflevector cases.
Craig Topper [Thu, 8 Aug 2013 07:38:55 +0000 (07:38 +0000)]
Optimize mask generation for one of the DAG combiner shufflevector cases.

llvm-svn: 187961

11 years agoPPC: Map frin to round() not nearbyint() and rint()
Hal Finkel [Thu, 8 Aug 2013 04:31:34 +0000 (04:31 +0000)]
PPC: Map frin to round() not nearbyint() and rint()

Making use of the recently-added ISD::FROUND, which allows for custom lowering
of round(), the PPC backend will now map frin to round(). Previously, we had
been using frin to lower nearbyint() (and rint() via some custom lowering to
handle the extra fenv flags requirements), but only in fast-math mode because
frin does not tie-to-even. Several users had complained about this behavior,
and this new mapping of frin to round is certainly more appropriate (and does
not require fast-math mode).

In effect, this reverts r178362 (and part of r178337, replacing the nearbyint
mapping with the round mapping).

llvm-svn: 187960

11 years agoRevert r187927.
Arnold Schwaighofer [Thu, 8 Aug 2013 03:06:24 +0000 (03:06 +0000)]
Revert r187927.

Zhihao Yuan: Replace operator& with addressof in reference_wrapper constructor.

It breaks a clang bootstrap.

llvm-svn: 187959

11 years agoSplit the deprecated increment bool warning into a sub-group of -Wdeprecated
Richard Trieu [Thu, 8 Aug 2013 03:05:52 +0000 (03:05 +0000)]
Split the deprecated increment bool warning into a sub-group of -Wdeprecated
so that it can be toggled independently of other deprecated warnings.

llvm-svn: 187958

11 years agoRevert r187935 "Support for double width characters."
Arnold Schwaighofer [Thu, 8 Aug 2013 02:19:56 +0000 (02:19 +0000)]
Revert r187935 "Support for double width characters."

It broke a public build bot.

llvm-svn: 187957

11 years agoRevert "coff also doesn't have a ReadOnlySection yet, (!)"
David Majnemer [Thu, 8 Aug 2013 01:50:52 +0000 (01:50 +0000)]
Revert "coff also doesn't have a ReadOnlySection yet, (!)"

This reverts commit r77814.

We were sticking global constants in the .data section instead of in the
.rdata section when emitting for COFF.

This fixes PR16831.

llvm-svn: 187956

11 years agoEmit an error for enum increments and decrements in C++ mode.
Richard Trieu [Thu, 8 Aug 2013 01:50:23 +0000 (01:50 +0000)]
Emit an error for enum increments and decrements in C++ mode.
Fixes PR16394.

llvm-svn: 187955

11 years agoReflow for loop.
Eric Christopher [Thu, 8 Aug 2013 01:41:05 +0000 (01:41 +0000)]
Reflow for loop.

llvm-svn: 187954

11 years agoBe more rigorous about the sizes of forms and attributes.
Eric Christopher [Thu, 8 Aug 2013 01:41:00 +0000 (01:41 +0000)]
Be more rigorous about the sizes of forms and attributes.

llvm-svn: 187953

11 years agocatch-undefined-behavior.cpp: Remove names that don't appear in every build.
Filipe Cabecinhas [Thu, 8 Aug 2013 01:24:29 +0000 (01:24 +0000)]
catch-undefined-behavior.cpp: Remove names that don't appear in every build.

llvm-svn: 187952

11 years agoUBSan: Fix naming for IR variables in catch-undef-behavior.cpp
Filipe Cabecinhas [Thu, 8 Aug 2013 01:18:31 +0000 (01:18 +0000)]
UBSan: Fix naming for IR variables in catch-undef-behavior.cpp

llvm-svn: 187951

11 years agoAdd three missing CMake variables that need to make it to the
Chandler Carruth [Thu, 8 Aug 2013 01:15:37 +0000 (01:15 +0000)]
Add three missing CMake variables that need to make it to the
LLVMConfig.cmake file that is (I think) used in the stand-alone Clang
build, and causing link errors there w.r.t. curses.

llvm-svn: 187950

11 years agoChanged isPrint for U+00AD SOFT HYPHEN to return true.
Alexander Kornienko [Thu, 8 Aug 2013 01:10:50 +0000 (01:10 +0000)]
Changed isPrint for U+00AD SOFT HYPHEN to return true.

Summary:
This is consistent with MacOSX implementation, and most terminals
actually display this character (checked on gnome-terminal, lxterminal, lxterm,
Terminal.app, iterm2). Actually, this is in line with the ISO Latin 1 standard
(ISO 8859-1), which defines it differently from the Unicode Standard. More
information here: http://www.cs.tut.fi/~jkorpela/shy.html

Reviewers: gribozavr, jordan_rose

CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1310

llvm-svn: 187949

11 years agoUBSan: Fix alignment checks emitted in downcasts.
Filipe Cabecinhas [Thu, 8 Aug 2013 01:08:17 +0000 (01:08 +0000)]
UBSan: Fix alignment checks emitted in downcasts.

Summary:
UBSan was checking for alignment of the derived class on the pointer to
the base class, before converting. With some class hierarchies, this could
generate false positives.

Added test-case.

llvm-svn: 187948

11 years agoForward resetColor() et al to the underlying stream.
Ted Kremenek [Thu, 8 Aug 2013 00:43:30 +0000 (00:43 +0000)]
Forward resetColor() et al to the underlying stream.

llvm-svn: 187947

11 years agoRevert "[objc-arc] Track if we encountered an additive overflow while computing ...
Michael Gottesman [Thu, 8 Aug 2013 00:41:18 +0000 (00:41 +0000)]
Revert "[objc-arc] Track if we encountered an additive overflow while computing {TopDown,BottomUp}PathCounts and do nothing if it occured."

This reverts commit r187941.

The commit was passing on my os x box, but it is failing on some non-osx
platforms. I do not have time to look into it now, so I am reverting and will
recommit after I figure this out.

llvm-svn: 187946

11 years agoclang-cl: Support the run-time selection options (/MD, /MT et al.)
Hans Wennborg [Thu, 8 Aug 2013 00:17:41 +0000 (00:17 +0000)]
clang-cl: Support the run-time selection options (/MD, /MT et al.)

These flags set some preprocessor macros and injects a dependency
on the runtime library into the object file, which later is picked up
by the linker.

This also adds a new CC1 flag for adding a dependent library.

Differential Revision: http://llvm-reviews.chandlerc.com/D1315

llvm-svn: 187945

11 years agoFix ARM build.
Peter Collingbourne [Thu, 8 Aug 2013 00:15:27 +0000 (00:15 +0000)]
Fix ARM build.

llvm-svn: 187944

11 years agoExpanded modularize tests.
John Thompson [Thu, 8 Aug 2013 00:01:09 +0000 (00:01 +0000)]
Expanded modularize tests.

llvm-svn: 187943

11 years agoComment changes. Added current problems comment.
John Thompson [Thu, 8 Aug 2013 00:00:10 +0000 (00:00 +0000)]
Comment changes.  Added current problems comment.

llvm-svn: 187942

11 years ago[objc-arc] Track if we encountered an additive overflow while computing {TopDown...
Michael Gottesman [Wed, 7 Aug 2013 23:56:41 +0000 (23:56 +0000)]
[objc-arc] Track if we encountered an additive overflow while computing {TopDown,BottomUp}PathCounts and do nothing if it occured.

rdar://14590914

llvm-svn: 187941

11 years ago[objc-arc] Change 4 iterator methods which return const_iterators to be const methods.
Michael Gottesman [Wed, 7 Aug 2013 23:56:34 +0000 (23:56 +0000)]
[objc-arc] Change 4 iterator methods which return const_iterators to be const methods.

llvm-svn: 187940

11 years agoReapply r185872 now that the address sanitizer has been changed to support this.
Bill Wendling [Wed, 7 Aug 2013 23:42:09 +0000 (23:42 +0000)]
Reapply r185872 now that the address sanitizer has been changed to support this.

Original commit message:

Stop emitting weak symbols into the "coal" sections.

The Mach-O linker has been able to support the weak-def bit on any symbol for
quite a while now. The compiler however continued to place these symbols into a
"coal" section, which required the linker to map them back to the base section
name.

Replace the sections like this:

  __TEXT/__textcoal_nt   instead use  __TEXT/__text
  __TEXT/__const_coal    instead use  __TEXT/__const
  __DATA/__datacoal_nt   instead use  __DATA/__data

<rdar://problem/14265330>

llvm-svn: 187939

11 years agoReally unbreak Mac build.
Peter Collingbourne [Wed, 7 Aug 2013 23:41:13 +0000 (23:41 +0000)]
Really unbreak Mac build.

llvm-svn: 187938

11 years agoUnbreak Mac build.
Peter Collingbourne [Wed, 7 Aug 2013 23:29:46 +0000 (23:29 +0000)]
Unbreak Mac build.

llvm-svn: 187937

11 years agoUpdate CREDITS.TXT
Howard Hinnant [Wed, 7 Aug 2013 23:29:10 +0000 (23:29 +0000)]
Update CREDITS.TXT

llvm-svn: 187936

11 years agoSupport for double width characters.
Alexander Kornienko [Wed, 7 Aug 2013 23:29:01 +0000 (23:29 +0000)]
Support for double width characters.

Summary: Only works for UTF-8-encoded files.

Reviewers: djasper

Reviewed By: djasper

CC: cfe-commits, klimek
Differential Revision: http://llvm-reviews.chandlerc.com/D1311

llvm-svn: 187935

11 years ago[lit] Remove unnecessary list copy.
Daniel Dunbar [Wed, 7 Aug 2013 23:10:20 +0000 (23:10 +0000)]
[lit] Remove unnecessary list copy.

llvm-svn: 187934

11 years ago[lit] Always list individual UNRESOLVED tests.
Daniel Dunbar [Wed, 7 Aug 2013 23:10:05 +0000 (23:10 +0000)]
[lit] Always list individual UNRESOLVED tests.

llvm-svn: 187933

11 years ago[lit] Explicitly convert dict items() result to a list.
Daniel Dunbar [Wed, 7 Aug 2013 23:10:01 +0000 (23:10 +0000)]
[lit] Explicitly convert dict items() result to a list.

llvm-svn: 187932

11 years ago[lit] Make string encoding issues explicit.
Daniel Dunbar [Wed, 7 Aug 2013 23:09:55 +0000 (23:09 +0000)]
[lit] Make string encoding issues explicit.

llvm-svn: 187931

11 years ago[tests] Make string encoding issues explicit.
Daniel Dunbar [Wed, 7 Aug 2013 23:09:24 +0000 (23:09 +0000)]
[tests] Make string encoding issues explicit.

llvm-svn: 187929

11 years ago[tests] Avoid deprecated except syntax.
Daniel Dunbar [Wed, 7 Aug 2013 23:09:10 +0000 (23:09 +0000)]
[tests] Avoid deprecated except syntax.

llvm-svn: 187928

11 years agoZhihao Yuan: Replace operator& with addressof in reference_wrapper constructor.
Howard Hinnant [Wed, 7 Aug 2013 23:02:42 +0000 (23:02 +0000)]
Zhihao Yuan: Replace operator& with addressof in reference_wrapper constructor.

llvm-svn: 187927

11 years agoAdd ISD::FROUND for libm round()
Hal Finkel [Wed, 7 Aug 2013 22:49:12 +0000 (22:49 +0000)]
Add ISD::FROUND for libm round()

All libm floating-point rounding functions, except for round(), had their own
ISD nodes. Recent PowerPC cores have an instruction for round(), and so here I'm
adding ISD::FROUND so that round() can be custom lowered as well.

For the most part, this is straightforward. I've added an intrinsic
and a matching ISD node just like those for nearbyint() and friends. The
SelectionDAG pattern I've named frnd (because ISD::FP_ROUND has already claimed
fround).

This will be used by the PowerPC backend in a follow-up commit.

llvm-svn: 187926

11 years agoDataFlowSanitizer; Clang changes.
Peter Collingbourne [Wed, 7 Aug 2013 22:47:34 +0000 (22:47 +0000)]
DataFlowSanitizer; Clang changes.

DataFlowSanitizer is a generalised dynamic data flow analysis.

Unlike other Sanitizer tools, this tool is not designed to detect a
specific class of bugs on its own.  Instead, it provides a generic
dynamic data flow analysis framework to be used by clients to help
detect application-specific issues within their own code.

Differential Revision: http://llvm-reviews.chandlerc.com/D966

llvm-svn: 187925

11 years agoDataFlowSanitizer; compiler-rt changes.
Peter Collingbourne [Wed, 7 Aug 2013 22:47:26 +0000 (22:47 +0000)]
DataFlowSanitizer; compiler-rt changes.

DataFlowSanitizer is a generalised dynamic data flow analysis.

Unlike other Sanitizer tools, this tool is not designed to detect a
specific class of bugs on its own.  Instead, it provides a generic
dynamic data flow analysis framework to be used by clients to help
detect application-specific issues within their own code.

Differential Revision: http://llvm-reviews.chandlerc.com/D967

llvm-svn: 187924

11 years agoDataFlowSanitizer; LLVM changes.
Peter Collingbourne [Wed, 7 Aug 2013 22:47:18 +0000 (22:47 +0000)]
DataFlowSanitizer; LLVM changes.

DataFlowSanitizer is a generalised dynamic data flow analysis.

Unlike other Sanitizer tools, this tool is not designed to detect a
specific class of bugs on its own.  Instead, it provides a generic
dynamic data flow analysis framework to be used by clients to help
detect application-specific issues within their own code.

Differential Revision: http://llvm-reviews.chandlerc.com/D965

llvm-svn: 187923

11 years agoFix doxygen function name mismatch.
Hans Wennborg [Wed, 7 Aug 2013 22:34:01 +0000 (22:34 +0000)]
Fix doxygen function name mismatch.

llvm-svn: 187922

11 years agoNew settings: target.use-hex-immediates and target.hex-immediates-style
Daniel Malea [Wed, 7 Aug 2013 21:54:09 +0000 (21:54 +0000)]
New settings: target.use-hex-immediates and target.hex-immediates-style
- Immediates can be shown as hex (either Intel or MASM style)
- See TestSettings.py for usage examples
- Verified to cause no regressions on Linux x86_64 (Ubuntu 12.10)

Patch by Richard Mitton!

llvm-svn: 187921

11 years ago[lit] Report the traceback when config import fails.
Daniel Dunbar [Wed, 7 Aug 2013 21:43:34 +0000 (21:43 +0000)]
[lit] Report the traceback when config import fails.

llvm-svn: 187920

11 years ago[lit] Avoid comparisons with None.
Daniel Dunbar [Wed, 7 Aug 2013 21:43:23 +0000 (21:43 +0000)]
[lit] Avoid comparisons with None.

llvm-svn: 187919

11 years ago[lit] Use list comprehensions instead of map().
Daniel Dunbar [Wed, 7 Aug 2013 21:43:17 +0000 (21:43 +0000)]
[lit] Use list comprehensions instead of map().

llvm-svn: 187918

11 years ago[lit] Avoid deprecated dict.has_key() method.
Daniel Dunbar [Wed, 7 Aug 2013 21:43:12 +0000 (21:43 +0000)]
[lit] Avoid deprecated dict.has_key() method.

llvm-svn: 187917

11 years agoPR9992: Serialize and deserialize the token sequence for a function template in
Richard Smith [Wed, 7 Aug 2013 21:41:30 +0000 (21:41 +0000)]
PR9992: Serialize and deserialize the token sequence for a function template in
-fdelayed-template-parsing mode. Patch by Will Wilson!

llvm-svn: 187916

11 years agoN3644 support for <unordered_set> and <unordered_map>
Marshall Clow [Wed, 7 Aug 2013 21:30:44 +0000 (21:30 +0000)]
N3644 support for <unordered_set> and <unordered_map>

llvm-svn: 187915

11 years ago[PCH] Fix a PCH serialization crash, with invalid code related to forward enum refere...
Argyrios Kyrtzidis [Wed, 7 Aug 2013 21:17:33 +0000 (21:17 +0000)]
[PCH] Fix a PCH serialization crash, with invalid code related to forward enum references.

The problem was that an enum without closing semicolon could be associated as a forward enum
in an erroneous declaration, leading to the identifier being associated with the enum decl but
without a declaration actually referencing it.
This resulted in not having it serialized before serializing the identifier that is associated with.

Also prevent the ASTUnit from querying the serialized DeclID for an invalid top-level decl; it may not
have been serialized.

rdar://14539667

llvm-svn: 187914