We don't want to report initialization-order bugs when a destructor of a global
variable accesses dynamically initialized global from another
(not necessarily initialized) module. We do this by intercepting __cxa_atexit and
registrering our own callback that unpoisons shadow for all dynamically initialized
global variables.

llvm-svn: 182637

llvm-svn: 182599

llvm-svn: 182598

"check-lsan" now runs both the tests from lib/lsan/tests and any lit
tests found under lib/lsan/lit_tests.

llvm-svn: 182583

llvm-svn: 182582

llvm-svn: 182580

[ASan] Introduce SymbolizerPrepareForSandboxing(), which is a no-op on every platform except Linux (because we don't support sandboxing anywhere else yet)
On Linux we pre-cache the value of readlink("/proc/self/exe"), so that it can be later used when the sandbox has been turned on.

llvm-svn: 182579

llvm-svn: 182578

llvm-svn: 182576

llvm-svn: 182575

llvm-svn: 182574

llvm-svn: 182573

Also remove unnecessary ifdefs.

llvm-svn: 182571

[ASan] Rename __asan_preinit to __local_asan_preinit to avoid warnings from the linker which is trying to export all the __asan_* symbols

llvm-svn: 182567

Using fwrite and fread was very *very* slow. The resulting code was multiple
times slower than GCC's implementation of gcov. Replace the fwrite/fread system
with an mmap() version.

If the `.gcda' file doesn't exist, we (re)allocate a buffer that we write
into. That gets written to the `.gcda' file in one chunk. If the `.gcda' file
already exists, we simply mmap() the file, modify the mapped data, and use
msync() to write the contents out to disk. It's much easier than implementing
our own buffering scheme, and we don't have to use fwrite's and fread's
buffering.

For those who are numbers-oriented, here are some timings:

GCC Verison
-----------

`.gcda' files don't exist:  23s
`.gcda' files do exist:     14s

LLVM Version (before this change)
---------------------------------

`.gcda' files don't exist:  28s
`.gcda' files do exist:     28s

LLVM Version (with this change)
-------------------------------

`.gcda' files don't exist:  18s
`.gcda' files do exist:      4s

It's a win-win-win-win-lose-win-win scenario!

<rdar://problem/13466086>

llvm-svn: 182563

llvm-svn: 182481

llvm-svn: 182479

Effectively revert r182453 - atomic.c may not be compiled if host compiler doesn't understand _Atomic

llvm-svn: 182478

llvm-svn: 182477

llvm-svn: 182475

llvm-svn: 182473

llvm-svn: 182470

llvm-svn: 182469

Multiple connected changes:
- Ignore reads from nested interceptors.
- Check shadow on reads from common interceptors.
- getaddrinfo interceptor.

llvm-svn: 182466

llvm-svn: 182465

llvm-svn: 182463

[asan] attempting to fix the debian bot (on ubuntu the test builds; on debian it complains that _SC_PAGESIZE is unknown)

llvm-svn: 182462

[asan] fix dlclose-test to work with non-4K pages. Properly disable interface_symbols_linux on non-x86

llvm-svn: 182461

[asan] workaround for asan bug 189 (swapcontext followed by throw gets OOM kill). Also, disable swapcontext_test on non-x86. Fix lint

llvm-svn: 182456

llvm-svn: 182453

llvm-svn: 182435

llvm-svn: 182420

Keeps it consistent between ASan and LSan.

llvm-svn: 182399

llvm-svn: 182389

llvm-svn: 182384

Also, define CAN_SANITIZE_LEAKS.

llvm-svn: 182383

llvm-svn: 182382

llvm-svn: 182380

llvm-svn: 182379

llvm-svn: 182378