Make memento checks more stable. Add filler at the end of new space and check if object and memento…

Make memento checks more stable. Add filler at the end of new space and check if object and memento are on the same new space page.

BUG=
R=bmeurer@chromium.org, mvstanton@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=19024

[Hannes Payer (out of office), 2014-02-03 10:30:18]

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h#newcode506
src/heap-inl.h:506: 
Not that I removed (ptr == top) since we have filler now all the time. I can
bring back this check, but it shouldn't matter anymore. WDYT?

[Benedikt Meurer, 2014-02-03 11:26:31]

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h#newcode491
src/heap-inl.h:491: ASSERT(heap->InFromSpace(object));
Is this ASSERT actually correct?

[Benedikt Meurer, 2014-02-03 11:49:28]

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h#newcode506
src/heap-inl.h:506: 
On 2014/02/03 10:30:18, Hannes Payer wrote:
> Not that I removed (ptr == top) since we have filler now all the time. I can
> bring back this check, but it shouldn't matter anymore. WDYT?

Please add a code comment why it's safe to skip the top/high check here.

[mvstanton, 2014-02-03 11:52:01]

lgtm with comment.

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h#newcode506
src/heap-inl.h:506: 
On 2014/02/03 10:30:18, Hannes Payer wrote:
> Not that I removed (ptr == top) since we have filler now all the time. I can
> bring back this check, but it shouldn't matter anymore. WDYT?

Indicate with a comment why you don't need this check.

[Hannes Payer (out of office), 2014-02-03 11:56:27]

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h
File src/heap-inl.h (right):

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h#newcode491
src/heap-inl.h:491: ASSERT(heap->InFromSpace(object));
On 2014/02/03 11:26:32, Benedikt Meurer wrote:
> Is this ASSERT actually correct?

Yes, because we should just look at objects which are in from space.

https://codereview.chromium.org/152613002/diff/20006/src/heap-inl.h#newcode506
src/heap-inl.h:506: 
On 2014/02/03 11:49:28, Benedikt Meurer wrote:
> On 2014/02/03 10:30:18, Hannes Payer wrote:
> > Not that I removed (ptr == top) since we have filler now all the time. I can
> > bring back this check, but it shouldn't matter anymore. WDYT?
> 
> Please add a code comment why it's safe to skip the top/high check here.

Done.

[Benedikt Meurer, 2014-02-03 11:56:45]

LGTM

[Hannes Payer (out of office), 2014-02-03 12:20:24]

Committed patchset #3 manually as r19024.