Drop OS::IsOutsideAllocatedSpace() and move the tracking to the MemoryAllocator.

Drop OS::IsOutsideAllocatedSpace() and move the tracking to the MemoryAllocator.

Instead of globally tracking allocated space limits, which was
not implemented properly anyway (i.e. lack of synchronization
on the reading side), track it per MemoryAllocator (that is
per heap/isolate).

In particular, avoid to call IsBadWritePtr() on Windows, it is
obsolete and Microsoft strongly discourages its usage.

R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=16542

[Benedikt Meurer, 2013-09-04 13:50:01]

Hey Michi,
This is the change we're talking about offline this noon. PTAL
-- Benedikt

[Michael Starzinger, 2013-09-04 15:49:24]

LGTM if all comments are addressed.

https://codereview.chromium.org/23903008/diff/1/src/platform-linux.cc
File src/platform-linux.cc (left):

https://codereview.chromium.org/23903008/diff/1/src/platform-linux.cc#oldcode508
src/platform-linux.cc:508: void OS::TearDown() {
AFAICT, this makes OS::TearDown empty on all platforms. Let's drop it
completely.

https://codereview.chromium.org/23903008/diff/1/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/23903008/diff/1/src/spaces.cc#newcode315
src/spaces.cc:315: UpdateAllocatedSpaceLimits(base,
reinterpret_cast<char*>(base) + size);
No need for a reinterpret_cast here, our Address typedef should allow for
byte-based arithmetic.

https://codereview.chromium.org/23903008/diff/1/src/spaces.h
File src/spaces.h (right):

https://codereview.chromium.org/23903008/diff/1/src/spaces.h#newcode1091
src/spaces.h:1091: return false;
This looks like debugging leftovers. Let's use the real boundary check as a
return value.

[Benedikt Meurer, 2013-09-05 08:06:46]

https://codereview.chromium.org/23903008/diff/1/src/platform-linux.cc
File src/platform-linux.cc (left):

https://codereview.chromium.org/23903008/diff/1/src/platform-linux.cc#oldcode508
src/platform-linux.cc:508: void OS::TearDown() {
On 2013/09/04 15:49:25, Michael Starzinger wrote:
> AFAICT, this makes OS::TearDown empty on all platforms. Let's drop it
> completely.

Done.

https://codereview.chromium.org/23903008/diff/1/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/23903008/diff/1/src/spaces.cc#newcode315
src/spaces.cc:315: UpdateAllocatedSpaceLimits(base,
reinterpret_cast<char*>(base) + size);
On 2013/09/04 15:49:25, Michael Starzinger wrote:
> No need for a reinterpret_cast here, our Address typedef should allow for
> byte-based arithmetic.

Done.

https://codereview.chromium.org/23903008/diff/1/src/spaces.h
File src/spaces.h (right):

https://codereview.chromium.org/23903008/diff/1/src/spaces.h#newcode1091
src/spaces.h:1091: return false;
On 2013/09/04 15:49:25, Michael Starzinger wrote:
> This looks like debugging leftovers. Let's use the real boundary check as a
> return value.

Ups, indeed.. :-)

[Benedikt Meurer, 2013-09-05 08:18:13]

Committed patchset #5 manually as r16542 (presubmit successful).