This corresponds to a fix in llvm::SplitBlockAndInsertIfThen.

llvm-svn: 204229

llvm-svn: 204228

In member function 'virtual void __sanitizer::DD::MutexBeforeLock(__sanitizer::DDCallback*, __sanitizer::DDMutex*, bool)':
error: the frame size of 544 bytes is larger than 512 bytes [-Werror=frame-larger-than=]

The code is now [arguably] better as well.

llvm-svn: 204227

[sanitizer] deadlock detector: a) initial support for suppressions, b) be more robust in case we failed to extract a stack trace for one of the locks

llvm-svn: 204225

Reviewed at http://llvm-reviews.chandlerc.com/D3107

llvm-svn: 204218

negative shift amounts and/or shifts wider than the type. VAX traps on
the former, X86 and other platforms produce incorrect results on the
latter.

llvm-svn: 204193

llvm-svn: 204185

llvm-svn: 204182

llvm-svn: 204179

If the user requests OS default stack size, do not adjust it to our minimum
stack size (which is usually much less than the OS default).

llvm-svn: 204173

llvm-svn: 204154

Compiler-rt part of MSan implementation of advanced origin tracking,
when we record not only creation point, but all locations where
an uninitialized value was stored to memory, too.

llvm-svn: 204152

the first flags is to enable printing of the second stack per edge

llvm-svn: 204150

llvm-svn: 204149

llvm-svn: 204148

llvm-svn: 204146

llvm-svn: 204145

this is necessaary because dlsym can call malloc, which can lock mutexes that we intercept

llvm-svn: 204141

llvm-svn: 204139

Google is re-licensing this code under the standard dual license of
compiler-rt.

llvm-svn: 204128

llvm-svn: 204127

still experimental

llvm-svn: 204126

llvm-svn: 204124

llvm-svn: 204119

llvm-svn: 204115

Reapply r204079 and r204083, this time with stubs for fputc in
compiler-rt.

llvm-svn: 204091

Buildbots are having trouble finding fputc(), and I can't figure out
why.  Reverting to investigate.

llvm-svn: 204088

llvm-svn: 204084

<rdar://problem/15943240>

llvm-svn: 204083

In instrumentation-based profiling, we need a set of data structures to
represent the counters.  Previously, these were built up during static
initialization.  Now, they're shoved into a specially-named section so
that they show up as an array.

As a consequence of the reorganizing symbols, instrumentation data
structures for linkonce functions are now correctly coalesced.

This is the first step in a larger project to minimize runtime overhead
and dependencies in instrumentation-based profilng.  The larger picture
includes removing all initialization overhead and making the dependency
on libc optional.

<rdar://problem/15943240>

llvm-svn: 204079

llvm-svn: 204044

[sanitizer] reverse the order of the stack traces printed for every pair of locks in the deadlock report (first print the 'from' node, then print the 'to' node of the deadlock graph)

llvm-svn: 204043

[sanitizer] make the deadlock detector print 2*N stack traces on lock-order-inversion with N locks (i.e. print stack traces for both lock acquisitions in every edge in the graph). More improvements to follow

llvm-svn: 204042

[sanitizer] allow to store the lock context (stack trace id) with all currently held locks in a thread. This will allow the deadlock detector to provide better warnings (more changes to go)

llvm-svn: 204039

llvm-svn: 204037

llvm-svn: 204036

llvm-svn: 204035

llvm-svn: 204034

Make behavior introduced in r202820 conditional (under legacy_pthread_cond flag).
The new issue that we've hit with the satellite pthread_cond_t struct is
that pthread_condattr_getpshared does not work (satellite data is not shared between processes).
The idea is that most processes do not use pthread 2.2.5.
The rare ones that use (2.2.5 is dated by 2002) must specify legacy_pthread_cond=1
on their own risk.

llvm-svn: 204032

llvm-svn: 204029