llvm-svn: 199811

llvm-svn: 199810

llvm-svn: 199809

llvm-svn: 199808

llvm-svn: 199807

llvm-svn: 199806

llvm-svn: 199805

llvm-svn: 199804

llvm-svn: 199803

implement correct atomic load/store for ARM
add test for atomic load/store
http://llvm-reviews.chandlerc.com/D2582

llvm-svn: 199802

llvm-svn: 199801

llvm-svn: 199800

Each functions is exported as "dllexport" in include/clang-c.
See also KB835326.

llvm-svn: 199799

because vector compare instruction puts result in mask register.

llvm-svn: 199798

MachineCopyPropagation has special logic for removing COPY instructions. It will remove plain COPYs using eraseFromParent(), but if the COPY has imp-defs/imp-uses it will convert it to a KILL, to keep the imp-def around.

This actually totally breaks and causes the machine verifier to cry in several cases, one of which being:

%RAX<def> = COPY %RCX<kill>
%ECX<def> = COPY %EAX<kill>, %RAX<imp-use,kill>

These subregister copies are together identified as noops, so are both removed. However, the second one as it has an imp-use gets converted into a kill:

%ECX<def> = KILL %EAX<kill>, %RAX<imp-use,kill>

As the original COPY has been removed, the verifier goes into tears at the use of undefined EAX and RAX.

There are several hacky solutions to this hacky problem (which is all to do with imp-use/def weirdnesses), but the least hacky I've come up with is to *always* remove COPYs by converting to KILLs. KILLs are no-ops to the code generator so the generated code doesn't change (which is why they were partially used in the first place), but using them also keeps the def/use and imp-def/imp-use chains alive:

%RAX<def> = KILL %RCX<kill>
%ECX<def> = KILL %EAX<kill>, %RAX<imp-use,kill>

The patch passes all test cases including the ones that check the removal of MOVs in this circumstance, along with an extra test I added to check subregister behaviour (which made the machine verifier fall over before my patch).

The patch also adds some DEBUG() statements because the file hadn't got any.

llvm-svn: 199797

Before:
  optional really.really.long.and.qualified.type.aaaaaaa
      .aaaaaaaa another_fiiiiiiiiiiiiiiiiiiiiield = 2;

After:
  optional
     really.really.long.and.qualified.type.aaaaaaa.aaaaaaaa
          another_fiiiiiiiiiiiiiiiiiiiiield = 2;

llvm-svn: 199796

It's guaranteed to be a CXXMethodDecl.

llvm-svn: 199795

Lift the getFunctionDecl() utility out of the parser into a general
Decl::getAsFunction() and use it to simplify other parts of the implementation.

Reduce isFunctionOrFunctionTemplate() to a simple type check that works the
same was as the other is* functions and move unwrapping of shadowed decls to
callers so it doesn't get run twice.

Shuffle around canSkipFunctionBody() to reduce virtual dispatch on ASTConsumer.
There's no need to query when we already know the body can't be skipped.

llvm-svn: 199794

This helped catch a couple of bugs locally.

llvm-svn: 199793

llvm-svn: 199791

This involved making CheckReturnStackAddr into a static function, which
is now called by a top-level return value checking routine called
CheckReturnValExpr.

llvm-svn: 199790

llvm-svn: 199788

We don't currently support this option, but we should be able to parse it.

llvm-svn: 199787

llvm-svn: 199786

llvm-svn: 199785

llvm-svn: 199784

llvm-svn: 199783

the program, in C++. (We allow the latter as an extension, since we've always
permitted it, and GCC does the same, and our supported C++ ABIs don't do
anything special in main.)

llvm-svn: 199782

llvm-svn: 199781

when the deployment target is 10.5. // rdar://15852259

llvm-svn: 199780

AST consumers can determine where the destructor name was written. Patch by
Olivier Goffart!

llvm-svn: 199779

language options. This is not really ideal -- we should require the right
language options to be passed in, or not require language options to format a
name -- but it fixes a number of *obviously* wrong formattings. Patch by
Olivier Goffart!

llvm-svn: 199778

llvm-svn: 199777

llvm-svn: 199776

Fixes PR#18521

llvm-svn: 199775

Checking in ActOnVariableDeclarator computes and caches the linkage using
the non-deduced auto type which defaults to external linkage. Depending on
how the auto type is deduced linkage can change and conflict with the
cached linkage, hitting asserts.

llvm-svn: 199774

Cleanup only.

llvm-svn: 199773

type if the capture is, actually, by copy.

llvm-svn: 199772

inconsistent results for different orderings of alloca slices. The
fundamental issue is that it is just always a mistake to return early
from this function. There is no effective early exit to leverage. This
patch stops trynig to do so and simplifies the code a bit as
a consequence.

Original diagnosis and patch by James Molloy with some name tweaks by me
in part reflecting feedback from Duncan Smith on the mailing list.

llvm-svn: 199771

The constructors of classes deriving from Binary normally take an error_code
as an argument to the constructor. My original intent was to change them
to have a trivial constructor and move the initial parsing logic to a static
method returning an ErrorOr. I changed my mind because:

* A constructor with an error_code out parameter is extremely convenient from
  the implementation side. We can incrementally construct the object and give
  up when we find an error.
* It is very efficient when constructing on the stack or when there is no
  error. The only inefficient case is where heap allocating and an error is
  found (we have to free the memory).

The result is that this is a much smaller patch. It just standardizes the
create* helpers to return an ErrorOr.

Almost no functionality change: The only difference is that this found that
we were trying to read past the end of COFF import library but ignoring the
error.

llvm-svn: 199770