When a linkonce_odr value that is on the dso list is not unnamed_addr
we can still look to see if anything is actually using its address. If
not, it is safe to hide it.

This patch implements that by moving GlobalStatus to Transforms/Utils
and using it in Internalize.

llvm-svn: 193090

Found while adding type safety to the various DWARF enumerations (form,
attribute, tag, etc) that caused Clang to warn on an incompletely
covered switch. Converting the comment to a default/unreachable
uncovered this case of an unsupported form encoding. Seems we were
skipping fission strings entirely.

llvm-svn: 193089

Patch by GM: Turn off 'deprecated' warnings when building with MSVC, and add '-Werror=return-type' to catch funtions that aren't returning what they should.

llvm-svn: 193088

Patch by GM: apparently '__value' (two underscores) is a special name in Visual Studio, so rename the private method in <regex> with that name. GM's patch used  '___value' (three underscores), but I changed that to '__regex_traits_value' because I've been burned in the past by identifiers that appear identical but are not.

llvm-svn: 193087

Patch from GM: locale.cpp; make implicit conversions to bool explicit, fix some 'unknown pragma' warnings when compiling under MSVC, and don't use the __sso_allocator under windows, b/c MSVC doesn't support aligned-by value parameters

llvm-svn: 193086

llvm-svn: 193085

llvm-svn: 193084

llvm-svn: 193083

llvm-svn: 193082

llvm-svn: 193081

llvm-svn: 193080

Also fixes some funky formatting.

llvm-svn: 193079

These instructions are logically related as they allow read/write of MSA control registers.
Currently MSA control registers are emitted by number but hopefully that will change as soon 
as GAS starts accepting them by name as that would make the assembly easier to read.

llvm-svn: 193078

llvm-svn: 193077

The second parameter of the SLD intrinsic is the number of columns (GPR) to 
slide left the source array.

llvm-svn: 193076

llvm-svn: 193075

Add the __ARM_ARCH_EXT_IDIV__ predefine. It is set to 1 if we have hardware divide in the mode that we are compiling in (depending on the target features), not defined if we don't. Should be compatible with the GCC conterpart. Also adding a -hwdiv option to overide the default behavior.

llvm-svn: 193074

Due to statement expressions supported as GCC extension, it is possible
to put 'break' or 'continue' into a loop/switch statement but outside its
body, for example:

    for ( ; ({ if (first) { first = 0; continue; } 0; }); )

Such usage must be diagnosed as an error, GCC rejects it. To recognize
this and similar patterns the flags BreakScope and ContinueScope are
temporarily turned off while parsing condition expression.

Differential Revision: http://llvm-reviews.chandlerc.com/D1762

llvm-svn: 193073

[asan] count the size of QuarantineBatch in the total Quarantine size; make QuarantineBatch fit into 8K, fix a MSVC compile warning

llvm-svn: 193072

Now that we iterate on the formatting multiple times when we
have chains of preprocessor branches, we need to correctly reset
the token's previous and next pointer for the first / last token.

llvm-svn: 193071

llvm-svn: 193070

Review: http://llvm-reviews.chandlerc.com/D1974
llvm-svn: 193069

Additionally some small comment/stylistic fixes are included as well.

llvm-svn: 193068

The C and C++ standards disallow using universal character names to
refer to some characters, such as basic ascii and control characters,
so we reject these sequences in the lexer. However, when the
preprocessor isn't being used on C or C++, it doesn't make sense to
apply these restrictions.

Notably, accepting these characters avoids issues with unicode escapes
when GHC uses the compiler as a preprocessor on haskell sources.

Fixes rdar://problem/14742289

llvm-svn: 193067

Avoid using the name 'bzero' for an enumerator in global scope. <strings.h> might declare this as a function.

llvm-svn: 193066

We claimed that the __is_interface_class keyword was KEYCXX even though
the __interface keyword was KEYMS.

llvm-svn: 193065

A landing pad can be jumped to only by the unwind edge of an invoke
instruction. If we eliminate a partially redundant load in a landing pad, it
will create a basic block that violates this constraint. It then leads to other
problems down the line if it tries to merge that basic block with the landing
pad. Avoid this by not eliminating the load in a landing pad.

PR17621

llvm-svn: 193064

llvm-svn: 193063

llvm-svn: 193062

llvm-svn: 193061

Differential Revision: http://llvm-reviews.chandlerc.com/D1339

llvm-svn: 193060

llvm-svn: 193059

This uses function prefix data to store function type information at the
function pointer.

Differential Revision: http://llvm-reviews.chandlerc.com/D1338

llvm-svn: 193058

llvm-svn: 193057

The convention is LF unless specifically testing line endings.

llvm-svn: 193056

ResolveSingleFunctionTemplateSpecialization() returns 0 and doesn't emit diags
unless the expression has template-ids, so we must null check the result.

Also add a better diag noting which overloads are causing the problem.

Reviewed by Aaron Ballman.

llvm-svn: 193055

Before:
  auto PointerBinding = [](const char * S) {};

After:
  auto PointerBinding = [](const char *S) {};

This fixes llvm.org/PR17618.

llvm-svn: 193054

llvm-svn: 193053

llvm-svn: 193052

Before:
  DEBUG({ // Comment that used to confuse clang-format.
  fdafas();
  });
Before:
  DEBUG({ // Comments are now fine.
    fdafas();
  });

This fixed llvm.org/PR17619.

llvm-svn: 193051