RegisterCoalescer set <undef> flags on all operands of copy instructions
that are scheduled to be removed. This is so they won't affect
shrinkToUses() by introducing false register reads.

Make sure those <undef> flags are never cleared, or shrinkToUses() could
cause live intervals to end at instructions about to be deleted.

This would be a lot simpler if RegisterCoalescer could just erase joined
copies immediately instead of keeping all the to-be-deleted instructions
around.

This fixes PR12862. Unfortunately, bugpoint can't create a sane test
case for this. Like many other coalescer problems, this failure depends
of a very fragile series of events.

<rdar://problem/11474428>

llvm-svn: 157001

Make sure useless (def-only) intervals also get verified.

llvm-svn: 157000

migration error.

This is more trouble that it is worth; autoreleasing a value without holding on it
is a valid use-case, we should not "punish" correct code for the minority of
broken/fragile programs that depend on the behavior of  -autorelease.

rdar://9914061

llvm-svn: 156999

bugpoint-ing, it may turn into something else.

llvm-svn: 156998

llvm-svn: 156996

separate side table, using the handy SequenceToOffsetTable class.  This encodes all
these weird things into another 256 bytes, allowing all intrinsics to be encoded this way.

llvm-svn: 156995

TestBackticksWithoutATarget.BackticksWithNoTargetTestCase was calling
GetDummyTarget() when executing for x86_64. When performing session
tearDown, it would get destroyed (and everything would be invalid (arch,
etc).

Then the test would run for i386. The dummy target wasn't being
reinitialized and was invalid. lldb complained that 'current process state
is unsuitable for expression parsing'.

llvm-svn: 156994

llvm-svn: 156993

tsan runtime shutdown is problematic for 2 reasons:
1. others crash during shutdown
2. we have to override user exit status (don't know it and can't return from atexit handler)

llvm-svn: 156991

http://codereview.appspot.com/6214052

llvm-svn: 156990

Patch by Meador Inge.

llvm-svn: 156989

llvm-svn: 156988

llvm-svn: 156987

llvm-svn: 156986

SelectionDAGBuilder: CaseBlock, CaseRanges and CaseCmp changed representation of Low and High from signed to unsigned. Since unsigned ints usually simpler, faster and allows to reduce some extra signed bit checks needed before <,>,<=,>= comparisons.

llvm-svn: 156985

llvm-svn: 156984

llvm-svn: 156983

llvm-svn: 156982

llvm-svn: 156981

for future expansion, no functionality change yet though.

llvm-svn: 156979

are only rejected because they can't be encoded into a 32-bit unit, not because
they contain an unencodable feature.

llvm-svn: 156978

intrinsics that use passed-in arguments.

llvm-svn: 156977

compatibility with LLVM 2.x bitcode files.

llvm-svn: 156976

llvm-svn: 156975

Make sure to subtract one from the PC when doing the symbolication of stack frames when it isn't the zero'th frame.

llvm-svn: 156974

llvm-svn: 156973

llvm-svn: 156971

about argument type mismatch.

This gives a nicer diagnostic in cases like
  printf(fmt,
         i);
where previously the snippet just pointed at 'fmt' (with a note at the
definition of fmt).

It's a wash for cases like
  printf("%f",
         i);
where previously we snippeted the offending portion of the format string,
but didn't indicate which argument was at fault.

llvm-svn: 156968

of the DeclStmt node, otherwise libclang will not work for anything inside that
class.

rdar://10837710

llvm-svn: 156966

used by the preprocessor.  Apple's GCC also supported a -A option for linking.

The ld man page has the following:
 -A basefile - Obsolete incremental load format.  This option is obsolete.

Nick Kledzik confirms this option is no longer needed/supported.
rdar://11455614

llvm-svn: 156965

Disambiguate past such a potential problem, and use the absence of 'typename'
to break ties in favor of a parenthesized thingy being an initializer, if
nothing else in the declaration disambiguates it as declaring a function.

llvm-svn: 156963

Added a testcase.

llvm-svn: 156961

TableGen already computes register units as the basic unit of
interference. We can use that to compute the set of overlapping
registers.

This means that we can easily compute overlap sets for one register at a
time. There is no benefit to computing all registers at once.

llvm-svn: 156960

llvm-svn: 156959

llc to recognize MIPS16 as a MIPS ASE extension. -mips16 will mean the
mips16 ASE for mips32 by default.

As part of fixing of adding this we discovered some small changes that
need to be made to MipsInstrInfo::storeRegToStackSLot and
MipsInstrInfo::loadRegFromStackSlot. We were using some "==" equality tests
where in fact we should have been using Mips::<regclas>.hasSubClassEQ instead,
per suggestion of Jakob Stoklund Olesen.

Patch by Reed Kotler.

llvm-svn: 156958

FIXME: JSON doesn't like path separator '\', on Win32 hosts.
llvm-svn: 156957

Make sure we can fail to create a compile unit without asserting. We now emit a warning.

llvm-svn: 156956

llvm-svn: 156955

Since r156650, clang has stopped emitting byval arguments for MIPS targets.

llvm-svn: 156954

llvm-svn: 156953