MemorySanitizer support.

MemorySanitizer support.

Enables building (including generating snapshot) and running simple scripts under MemorySanitizer.
Not all tests pass yet.

R=asiva@google.com

Committed: https://code.google.com/p/dart/source/detail?r=42561

[siva, 2014-12-22 17:01:53]

LGTM

I would prefer if we used macros
MSAN_UNPOISON(...)
ASAN_.....

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/address...
File runtime/platform/address_sanitizer.h (right):

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/address...
runtime/platform/address_sanitizer.h:19: static void
__asan_unpoison_memory_region(void* ignore1, size_t ignore2) {}
Will these always be optimized out on all systems and compilers,
why not use macros that expand to nothing (we do that in a lot of places e.g
NoGcScope, NoHandleScope etc.

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/memory_...
File runtime/platform/memory_sanitizer.h (right):

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/memory_...
runtime/platform/memory_sanitizer.h:19: static void __msan_unpoison(void*
ignore1, size_t ignore2) {}
Ditto comment about using macro instead of an empty function like this.

https://codereview.chromium.org/816123002/diff/60001/runtime/vm/stack_frame.cc
File runtime/vm/stack_frame.cc (right):

https://codereview.chromium.org/816123002/diff/60001/runtime/vm/stack_frame.c...
runtime/vm/stack_frame.cc:284: static void UnpoisonStack(Isolate* isolate) {
ASSERT(isolate->stack_limit() != ~static_cast<uword>(0));

[koda, 2014-12-22 23:05:52]

Added macros and changed UnpoisonStack to not rely on stack limit being valid.

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/address...
File runtime/platform/address_sanitizer.h (right):

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/address...
runtime/platform/address_sanitizer.h:19: static void
__asan_unpoison_memory_region(void* ignore1, size_t ignore2) {}
On 2014/12/22 17:01:53, siva wrote:
> Will these always be optimized out on all systems and compilers,
> why not use macros that expand to nothing (we do that in a lot of places e.g
> NoGcScope, NoHandleScope etc.

Done.

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/memory_...
File runtime/platform/memory_sanitizer.h (right):

https://codereview.chromium.org/816123002/diff/60001/runtime/platform/memory_...
runtime/platform/memory_sanitizer.h:19: static void __msan_unpoison(void*
ignore1, size_t ignore2) {}
On 2014/12/22 17:01:53, siva wrote:
> Ditto comment about using macro instead of an empty function like this.

Done.

https://codereview.chromium.org/816123002/diff/60001/runtime/vm/stack_frame.cc
File runtime/vm/stack_frame.cc (right):

https://codereview.chromium.org/816123002/diff/60001/runtime/vm/stack_frame.c...
runtime/vm/stack_frame.cc:284: static void UnpoisonStack(Isolate* isolate) {
On 2014/12/22 17:01:53, siva wrote:
> ASSERT(isolate->stack_limit() != ~static_cast<uword>(0));

Changed to instead take an explicit 'fp' argument, and unpoison the area around
it.

[siva, 2014-12-22 23:39:57]

still LGTM

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/exceptions.cc
File runtime/vm/exceptions.cc (right):

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/exceptions.c...
runtime/vm/exceptions.cc:303: stack_pointer - current_sp);
indent problem?

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/native_entry.h
File runtime/vm/native_entry.h (right):

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/native_entry...
runtime/vm/native_entry.h:48: MSAN_UNPOISON(arguments, sizeof(*arguments));     
                      \
indent issue on '  \'

[koda, 2014-12-23 01:07:42]

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/exceptions.cc
File runtime/vm/exceptions.cc (right):

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/exceptions.c...
runtime/vm/exceptions.cc:303: stack_pointer - current_sp);
On 2014/12/22 23:39:57, siva wrote:
> indent problem?

Done.

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/native_entry.h
File runtime/vm/native_entry.h (right):

https://codereview.chromium.org/816123002/diff/100001/runtime/vm/native_entry...
runtime/vm/native_entry.h:48: MSAN_UNPOISON(arguments, sizeof(*arguments));     
                      \
On 2014/12/22 23:39:57, siva wrote:
> indent issue on '  \'

Done.

[koda, 2014-12-23 01:08:05]

Committed patchset #7 (id:120001) manually as r42561 (presubmit successful).