with a suitable 'getter' attribute.

llvm-svn: 190420

It was removed in r189130, but it turns out this makes life hard for
folks packaging LLVM and Clang and building the latter based on the
LLVM package.

Note that this only adds back the LLVM tblgen, and it's obviously
not included when LLVM_INSTALL_TOOLCHAIN_ONLY is set.

llvm-svn: 190419

Specialize the constructors for DIRef<DIScope> and DIRef<DIType> to make sure
the Value is indeed a scope ref and a type ref.

Use DIScopeRef for DIScope::getContext and DIType::getContext and use DITypeRef
for getContainingType and getClassType.

DIScope::generateRef now returns a DIScopeRef instead of a "Value *" for
readability and type safety.

llvm-svn: 190418

We were figuring out whether to use tPICADD or PICADD, then just using
tPICADD unconditionally anyway. Oops.

A testcase from someone familiar enough with ELF to produce one would
be appreciated. The existing PIC testcase correctly verifies the .s
generated, but that doesn't catch this bug, which only showed up in
direct-to-object mode.

http://llvm.org/bugs/show_bug.cgi?id=17180

llvm-svn: 190417

llvm.org/pr17184 expression interpreter fails for crash/assert tests

llvm-svn: 190416

http://llvm.org/pr17183 expression w/ varargs printf() fails
http://llvm.org/pr15302 'anonymous namespace' prefix missing

llvm-svn: 190415

Fixup for r190410: use lazy initialization for symbolizer as some compilers emit global constructor to setup vptr

llvm-svn: 190414

Dotfiles are impractical on Windows. This makes clang-format search
for the style configuration file as '_clang-format' in addition to
the usual '.clang-format'. This is similar to how VIM searches for
'_vimrc' on Windows.

Differential Revision: http://llvm-reviews.chandlerc.com/D1629

llvm-svn: 190413

This commit removes the unused "AttributeItem" from
ObjectAttributeEmitter.

llvm-svn: 190412

llvm-svn: 190411

[Sanitizer] Refactor symbolization interface: use class instead of several functions. Move some code around to get rid of extra source files

llvm-svn: 190410

LibXML2 config doesn't specify lzma as a dependency, which breaks
cross-compilation builds using new linkers (ld 2.21 or higher).

There is a bug on libxml2 to fix that, but since it's going to take
a while for things to go round and back, so we should have a harmless
addition of the library until then.

llvm-svn: 190409

Reviewers: djasper

Reviewed By: djasper

CC: cfe-commits, klimek

Differential Revision: http://llvm-reviews.chandlerc.com/D1639

llvm-svn: 190408

[asan] refactor the use-after-return API so that the size class is computed at compile time instead of at run-time. llvm part

llvm-svn: 190407

[asan] refactor the use-after-return API so that the size class is computed at compile time instead of at run-time. compiler-rt part

llvm-svn: 190406

Summary: This also unifies the handling of escaped newlines for all tokens.

Reviewers: djasper

Reviewed By: djasper

CC: cfe-commits, klimek

Differential Revision: http://llvm-reviews.chandlerc.com/D1638

llvm-svn: 190405

llvm-svn: 190404

This way msan annotations can be used with both normal and volatile memory.

llvm-svn: 190403

llvm-svn: 190402

Before:
  typedef typeof(int(int, int)) * MyFunc;
After:
  typedef typeof(int(int, int)) *MyFunc;

This fixes llvm.org/PR17178.

llvm-svn: 190401

The main complication here is that TM and TMY (the memory forms) set
CC differently from the register forms.  When the tested bits contain
some 0s and some 1s, the register forms set CC to 1 or 2 based on the
value the uppermost bit.  The memory forms instead set CC to 1
regardless of the uppermost bit.

Until now, I've tried to make it so that a branch never tests for an
impossible CC value.  E.g. NR only sets CC to 0 or 1, so branches on the
result will only test for 0 or 1.  Originally I'd tried to do the same
thing for TM and TMY by using custom matching code in ISelDAGToDAG.
That ended up being very ugly though, and would have meant duplicating
some of the chain checks that the common isel code does.

I've therefore gone for the simpler alternative of adding an extra
operand to the TM DAG opcode to say whether a memory form would be OK.
This means that the inverse of a "TM;JE" is "TM;JNE" rather than the
more precise "TM;JNLE", just like the inverse of "TMLL;JE" is "TMLL;JNE".
I suppose that's arguably less confusing though...

llvm-svn: 190400

The dotp_[su].b instructions never existed in any revision of the MSA spec.

llvm-svn: 190399

The dotp_[su].b instructions never existed in any revision of the MSA spec.

llvm-svn: 190398

Add test cases for Mips mthc1/mfhc1 instructions. Add check for odd value of register when PFU is 32 bit.

llvm-svn: 190397

llvm-svn: 190396

Summary:
This fixes various issues with mixed tabs and spaces handling, e.g.
when realigning block comments.

Reviewers: klimek, djasper

Reviewed By: djasper

CC: cfe-commits

Differential Revision: http://llvm-reviews.chandlerc.com/D1608

llvm-svn: 190395

llvm-svn: 190394

llvm-svn: 190393

Also make DescribeThread easier to use.
This was firing on Dr.ASan runs, not sure how to repro InvalidTID in a simple test.

llvm-svn: 190392

PR16752: 'mode' attribute for unusual targets doesn't work properly
Description:
Troubles could be happened due to some assumptions in handleModeAttr function (see SemaDeclAttr.cpp).
For example, it assumes that 32 bit integer is 'int', while it could be 16 bit only.
Instead of asking target: 'which type do you want to use for int32_t ?' it just hardcodes general opinion. That doesn't looks pretty correct.
Please consider the next solution:
1. In Basic/TargetInfo add getIntTypeByWidth and getRealTypeByWidth methods. Methods asks target for proper type for given bit width.
2. Fix handleModeAttr according to new methods in TargetInfo.

Fixes:
1st Commit (Done): Add new methods for TargetInfo:
     getRealTypeByWidth and getIntTypeByWidth
  for ASTContext names are almost same(invokes new methods from TargetInfo):
     getIntTypeForBitwidth and getRealTypeForBitwidth

2nd Commit (Current): Fix SemaDeclAttr, handleModeAttr function.

llvm-svn: 190391

llvm-svn: 190390

[python-bindings] Changed test_memory_buffer_create_from_file to just use the generic provided test_file instead of a binary.

llvm-svn: 190389

This is a part of a series of patches that have been sitting fallow on a
personal branch that I have been messing with for a bit.

The patches start to flesh out the python llvm-c wrapper to the point where you can:

1. Load Modules from Bitcode/Dump/Print them.
2. Iterate over Functions from those modules/get their names/dump them.
3. Iterate over the BasicBlocks from said function/get the BB's name/dump it.
4. Iterate over the Instructions in said BasicBlocks/get the instructions
   name/dump the instruction.

My main interest in developing this was to be able to gather statistics about
LLVM IR using python scripts to speed up statistical profiling of different IR
level transformations (hence the focus on printing/dumping/getting names).

This is a gift from me to the LLVM community = ).

I am going to be committing the patches slowly over the next bit as I have time
to prepare the patches.

The overall organization follows the c-api like the bindings that are already
implemented.

llvm-svn: 190388

Enabling sse4.2 will implicitly enable popcnt unless popcnt is explicitly disabled.
Disabling sse4.2 will not disable popcnt if popcnt is explicitly enabled.

llvm-svn: 190387

XFAIL can be removed. (in r190374)

llvm-svn: 190386

for the copy assignment operator.

llvm-svn: 190385

Use 0 >= 1 instead of 0 != 0 to represent 'false'. This might be slightly more
efficient as isl may create a union of sets for 0 != 0, whereas this is never
needed for the expression 0 >= 1.

Contributed-by: Alexandre Isoard <alexandre.isoard@gmail.com>
llvm-svn: 190384

llvm-svn: 190383

This patch does a few different things.

This patch improves unused var diags for const vars: we no longer
unconditionally suppress diagnostics for const vars, instead only suppressing
the diagnostic when the declaration appears to be useful.

This patch also makes us more consistently use whether a variable/function
is declared in the main file to suppress diagnostics where appropriate.

Fixes <rdar://problem/14907887>.

llvm-svn: 190382

that /bin/sh re-exec's itself to /bin/bash, so it needs one more resume when you
are using it as the shell than /bin/bash did or you will stop at the start of your
program, rather than running it.

So I added a Platform API to get the number of resumes needed when launching with
a particular shell, and set the right values for Mac OS X.

<rdar://problem/14935282>

llvm-svn: 190381