Enable universal accelerated overflow scroll

Enable universal accelerated overflow scroll

Now that we've enabled layer squashing, we can enable universal overflow
scrolling.

Note: This is just an updated version of the patch abarth@ kindly
started here: https://codereview.chromium.org/243713002/

I've removed some now meaningless layout tests (which cover the old,
opt-in path), and updated the TestExpectaions for the remaining,
affected tests.

R=abarth@chromium.org
BUG=249354
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=172041

[abarth-chromium, 2014-04-19 21:53:48]

LGTM

https://codereview.chromium.org/244193002/diff/1/LayoutTests/TestExpectations
File LayoutTests/TestExpectations (right):

https://codereview.chromium.org/244193002/diff/1/LayoutTests/TestExpectations...
LayoutTests/TestExpectations:1139: crbug.com/364991 [ Debug ]
virtual/gpu/compositedscrolling/overflow/iframe-scroll-children.html [ Pass
Crash ]
Does this test still crash after your CL?  (Just curious)

[Ian Vollick, 2014-04-19 23:30:43]

On 2014/04/19 21:53:48, abarth wrote:
> LGTM
> 
> https://codereview.chromium.org/244193002/diff/1/LayoutTests/TestExpectations
> File LayoutTests/TestExpectations (right):
> 
>
https://codereview.chromium.org/244193002/diff/1/LayoutTests/TestExpectations...
> LayoutTests/TestExpectations:1139: crbug.com/364991 [ Debug ]
> virtual/gpu/compositedscrolling/overflow/iframe-scroll-children.html [ Pass
> Crash ]
> Does this test still crash after your CL?  (Just curious)

Sadly, yes. :(

There were some more opt-in layout tests that I'd missed earlier that caused the
bot redness. I think I've got the last of them as well as some code to support
them in Internals. Here's hoping the trybots are happier with this latest patch
set.

[abarth-chromium, 2014-04-20 02:24:33]

The CQ bit was checked by abarth@chromium.org

[commit-bot: I haz the power, 2014-04-20 02:24:44]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/vollick@chromium.org/244193002/30079

[commit-bot: I haz the power, 2014-04-20 03:02:54]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2014-04-20 03:02:54]

Try jobs failed on following builders:
  tryserver.blink on mac_blink_rel

[abarth-chromium, 2014-04-20 14:16:39]

Looks like you've got one more test:

virtual/softwarecompositing/overflow/universal-accelerated-overflow-scroll.html

[Ian Vollick, 2014-04-21 02:09:12]

On 2014/04/20 14:16:39, abarth wrote:
> Looks like you've got one more test:
> 
>
virtual/softwarecompositing/overflow/universal-accelerated-overflow-scroll.html

I seem to be running out of file descriptors. I appear to be allocating around
109, mostly tiny (smaller than a tile), shared bitmaps and that exceeds a
threshold. In the test, I validate 12 configurations, and when I test locally,
if I only try 8 of the configurations (it doesn't seem to matter which 8), the
test succeeds, more than that and I hit the limit. This might be
https://code.google.com/p/chromium/issues/detail?id=351897 ? 

I'm a bit torn about what the right thing to do is, here. I could batch the
configurations, but I'm not sure the demands I'm placing on the system are
really unusual (they might be, though; I'll need to ask), so I worry that
working around this issue would be papering over a nice repro case.

I could also mark the test as crashing for mac + software-compositing and cite
the bug.

wdyt?

[Ian Vollick, 2014-04-21 15:00:02]

On 2014/04/21 02:09:12, Ian Vollick wrote:
> On 2014/04/20 14:16:39, abarth wrote:
> > Looks like you've got one more test:
> > 
> >
>
virtual/softwarecompositing/overflow/universal-accelerated-overflow-scroll.html
> 
> I seem to be running out of file descriptors. I appear to be allocating around
> 109, mostly tiny (smaller than a tile), shared bitmaps and that exceeds a
> threshold. In the test, I validate 12 configurations, and when I test locally,
> if I only try 8 of the configurations (it doesn't seem to matter which 8), the
> test succeeds, more than that and I hit the limit. This might be
> https://code.google.com/p/chromium/issues/detail?id=351897 ? 
> 
> I'm a bit torn about what the right thing to do is, here. I could batch the
> configurations, but I'm not sure the demands I'm placing on the system are
> really unusual (they might be, though; I'll need to ask), so I worry that
> working around this issue would be papering over a nice repro case.
> 
> I could also mark the test as crashing for mac + software-compositing and cite
> the bug.
> 
> wdyt?

I'm leaning towards marking the test as crashing for mac + software-compositing.
I've uploaded a patch set to do this. I should mention that the reason this test
wasn't crashing before the change is that the settings were incorrect; we had
been running the legacy composited scrolling path. Switching the test to use the
new compositing path reproduces the crash on ToT with mac +
software-compositing.

[abarth-chromium, 2014-04-21 16:23:35]

Ok...  Do you think this issue can occur in the real world?

[abarth-chromium, 2014-04-21 16:23:44]

The CQ bit was checked by abarth@chromium.org

[commit-bot: I haz the power, 2014-04-21 16:23:54]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/vollick@chromium.org/244193002/90001

[commit-bot: I haz the power, 2014-04-21 16:34:23]

Change committed as 172041

[Ian Vollick, 2014-04-21 17:09:46]

On 2014/04/21 16:23:35, abarth wrote:
> Ok...  Do you think this issue can occur in the real world?

I think so, yeah. There's a good chance I'm misunderstanding the issue, but it
looks like on mac + software compositing, any situation that could create a
largish number of layers (and shared bitmaps) will cause us to exceed the file
descriptor limit.