Reland main thread position sticky implementation.

Reland main thread position sticky implementation.

Fix the crash and reland the position sticky implementation reviewed in https://codereview.chromium.org/1308273010. TBRing reviewers on unchanged files.

BUG=231752
TEST=fast/css/sticky/*
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
TBR=tdresser,isherman

Committed: https://crrev.com/86b31f0403a7bd7b7a768314b7ba9565af0c9a29
Cr-Commit-Position: refs/heads/master@{#386998}

[flackr, 2016-04-07 14:35:46]

Description was changed from

==========
Reland main thread position sticky implementation.

Fix the crash and reland the position sticky implementation reviewed in
https://codereview.chromium.org/1308273010.

BUG=231752
TEST=fast/css/sticky/*
==========

to

==========
Reland main thread position sticky implementation.

Fix the crash and reland the position sticky implementation reviewed in
https://codereview.chromium.org/1308273010.

BUG=231752
TEST=fast/css/sticky/*
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[flackr, 2016-04-07 21:43:53]

flackr@chromium.org changed reviewers:
+ eae@chromium.org

[flackr, 2016-04-07 21:43:54]

Hi Emil, would you mind reviewing patchset 2? It looks like Chris is away for a
little while - but he reviewed patchset 1 (It's a revert of the revert from when
it originally landed). There was a crash resulting from an invalid
ancestorOverflowLayer pointer which I have added a test for and fixed. Thanks!

[eae, 2016-04-11 23:40:43]

LGTM

[Ian Vollick, 2016-04-12 01:36:17]

On 2016/04/11 23:40:43, eae wrote:
> LGTM

cc/ still lgtm.

[flackr, 2016-04-12 14:26:23]

flackr@chromium.org changed reviewers:
+ vollick@chromium.org

[flackr, 2016-04-12 14:26:25]

Emil, see my comment inline, it may still be possible to crash if there is
another way for removed layer trees to be reattached besides
removeOnlyThisLayer. Let me know if you think we should be safe and just always
walk until we find a different ancestor overflow pointer in the tree. Thanks.

https://codereview.chromium.org/1870663002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/core/paint/PaintLayer.cpp (right):

https://codereview.chromium.org/1870663002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/core/paint/PaintLayer.cpp:1242:
oldChild->updateAncestorOverflowLayer(nullptr);
In order to ensure we don't have any stale ancestor pointers, the correct thing
to do would be to call removeAncestorOverflowLayer here, rather than in
removeOnlyThisLayer, but it may be wasteful to do this if the only time we
reattach the children layers is in removeOnlyThisLayer. Do you think we should
try the patch as is? Do you know of other cases where we reattach the child
branches that may also cause the crash?

[eae, 2016-04-12 15:48:15]

I'd argue for correctness over an early optimization.

[flackr, 2016-04-12 20:41:10]

On 2016/04/12 at 15:48:15, eae wrote:
> I'd argue for correctness over an early optimization.

Done. It now removes the ancestorOverflowLayer pointers which are not part of
the removed tree on removeLayer.

[flackr, 2016-04-12 20:44:35]

Description was changed from

==========
Reland main thread position sticky implementation.

Fix the crash and reland the position sticky implementation reviewed in
https://codereview.chromium.org/1308273010.

BUG=231752
TEST=fast/css/sticky/*
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Reland main thread position sticky implementation.

Fix the crash and reland the position sticky implementation reviewed in
https://codereview.chromium.org/1308273010. TBRing reviewers on unchanged files.

BUG=231752
TEST=fast/css/sticky/*
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
TBR=tdresser,isherman
==========

[flackr, 2016-04-12 20:44:35]

flackr@chromium.org changed reviewers:
+ isherman@chromium.org, tdresser@chromium.org

[flackr, 2016-04-13 14:08:14]

The CQ bit was checked by flackr@chromium.org

[flackr, 2016-04-13 14:08:15]

The patchset sent to the CQ was uploaded after l-g-t-m from
vollick@chromium.org, eae@chromium.org
Link to the patchset: https://codereview.chromium.org/1870663002/#ps60001
(title: "Only descend into children which have an ancestor overflow layer.")

[commit-bot: I haz the power, 2016-04-13 14:08:31]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1870663002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1870663002/60001

[commit-bot: I haz the power, 2016-04-13 15:42:01]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2016-04-13 15:43:11]

Description was changed from

==========
Reland main thread position sticky implementation.

Fix the crash and reland the position sticky implementation reviewed in
https://codereview.chromium.org/1308273010. TBRing reviewers on unchanged files.

BUG=231752
TEST=fast/css/sticky/*
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
TBR=tdresser,isherman
==========

to

==========
Reland main thread position sticky implementation.

Fix the crash and reland the position sticky implementation reviewed in
https://codereview.chromium.org/1308273010. TBRing reviewers on unchanged files.

BUG=231752
TEST=fast/css/sticky/*
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
TBR=tdresser,isherman

Committed: https://crrev.com/86b31f0403a7bd7b7a768314b7ba9565af0c9a29
Cr-Commit-Position: refs/heads/master@{#386998}
==========

[commit-bot: I haz the power, 2016-04-13 15:43:13]

Patchset 4 (id:??) landed as
https://crrev.com/86b31f0403a7bd7b7a768314b7ba9565af0c9a29
Cr-Commit-Position: refs/heads/master@{#386998}