Roll skia 2051->2062

Roll skia 2051->2062

Whitespace change: http://codereview.chromium.org/7600002

BUG=NONE
TEST=NONE

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=95874
Reverted: http://src.chromium.org/viewvc/chrome?view=rev&revision=95913

[vandebo (ex-Chrome), 2011-08-08 20:17:16]

The win layout bots still seem to be broken, see
http://build.chromium.org/p/tryserver.chromium/builders/win_layout

[bsalomon, 2011-08-08 20:29:55]

On 2011/08/08 20:17:16, vandebo wrote:
> The win layout bots still seem to be broken, see
> http://build.chromium.org/p/tryserver.chromium/builders/win_layout

LGTM

[vandebo (ex-Chrome), 2011-08-09 00:14:05]

The webkit sheriff asked me to revert this change because the WK canary started
to fail. These tests crashed on Linux: 2d.path.arc.zero.1.html
2d.path.bezierCurveTo.basic.html 2d.path.bezierCurveTo.shape.html
2d.path.fill.closed.unaffected.html 2d.path.fill.winding.subtract.3.html
2d.path.moveTo.basic.html 2d.path.rect.end.1.html
2d.path.rect.selfintersect.html 2d.path.rect.zero.4.html
2d.path.stroke.prune.arc.html 2d.path.stroke.prune.rect.html
2d.path.stroke.union.html 2d.shadow.canvas.basic.html 2d.shadow.enable.x.html
2d.state.saverestore.globalCompositeOperation.html
context.unrecognised.nullsuffix.html canvas-copyPixels.html
canvas-drawImage-incomplete.html canvas-fillPath-pattern-shadow.html
canvas-font-ex-units-crash.html

http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Linux/builds/11653

Any idea Brian?

[bsalomon, 2011-08-09 12:21:40]

Hmm... having a look. I did run the bots on my windows machine yesterday at
2061 and didn't get any unexpected failures.

On Mon, Aug 8, 2011 at 8:14 PM, <vandebo@chromium.org> wrote:

> The webkit sheriff asked me to revert this change because the WK canary
> started
> to fail. These tests crashed on Linux: 2d.path.arc.zero.1.html
> 2d.path.bezierCurveTo.basic.**html 2d.path.bezierCurveTo.shape.**html
> 2d.path.fill.closed.**unaffected.html 2d.path.fill.winding.subtract.**
> 3.html
> 2d.path.moveTo.basic.html 2d.path.rect.end.1.html
> 2d.path.rect.selfintersect.**html 2d.path.rect.zero.4.html
> 2d.path.stroke.prune.arc.html 2d.path.stroke.prune.rect.html
> 2d.path.stroke.union.html 2d.shadow.canvas.basic.html
> 2d.shadow.enable.x.html
> 2d.state.saverestore.**globalCompositeOperation.html
> context.unrecognised.**nullsuffix.html canvas-copyPixels.html
> canvas-drawImage-incomplete.**html canvas-fillPath-pattern-**shadow.html
> canvas-font-ex-units-crash.**html
>
> http://build.chromium.org/p/**chromium.webkit/builders/**
>
Webkit%20Linux/builds/11653<http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Linux/builds/11653>
>
> Any idea Brian?
>
>
>
http://codereview.chromium.**org/7600001/<http://codereview.chromium.org/7600...
>

[vandebo (ex-Chrome), 2011-08-09 17:09:01]

I suspect something changed in Webkit (maybe something was reverted), because
the try jobs from before I committed were clean, but the try on patch set 2 -
after the revert have the problem.

On 2011/08/09 12:21:40, bsalomon wrote:
> Hmm... having a look. I did run the bots on my windows machine yesterday at
> 2061 and didn't get any unexpected failures.
> 
> On Mon, Aug 8, 2011 at 8:14 PM, <mailto:vandebo@chromium.org> wrote:
> 
> > The webkit sheriff asked me to revert this change because the WK canary
> > started
> > to fail. These tests crashed on Linux: 2d.path.arc.zero.1.html
> > 2d.path.bezierCurveTo.basic.**html 2d.path.bezierCurveTo.shape.**html
> > 2d.path.fill.closed.**unaffected.html 2d.path.fill.winding.subtract.**
> > 3.html
> > 2d.path.moveTo.basic.html 2d.path.rect.end.1.html
> > 2d.path.rect.selfintersect.**html 2d.path.rect.zero.4.html
> > 2d.path.stroke.prune.arc.html 2d.path.stroke.prune.rect.html
> > 2d.path.stroke.union.html 2d.shadow.canvas.basic.html
> > 2d.shadow.enable.x.html
> > 2d.state.saverestore.**globalCompositeOperation.html
> > context.unrecognised.**nullsuffix.html canvas-copyPixels.html
> > canvas-drawImage-incomplete.**html canvas-fillPath-pattern-**shadow.html
> > canvas-font-ex-units-crash.**html
> >
> > http://build.chromium.org/p/**chromium.webkit/builders/**
> >
>
Webkit%20Linux/builds/11653<http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Linux/builds/11653>
> >
> > Any idea Brian?
> >
> >
> >
>
http://codereview.chromium.**org/7600001/%3Chttp://codereview.chromium.org/76...>
> >

[bsalomon, 2011-08-09 17:16:15]

I thinkr 2061 introduced a bug and changes in WebKit exposed it. There can
be a crash at GrContext teardown time. I've reproduced it locally by pulling
r2061 or skia and TOT WebKit and am working on a fix.

On Tue, Aug 9, 2011 at 1:09 PM, <vandebo@chromium.org> wrote:

> I suspect something changed in Webkit (maybe something was reverted),
> because
> the try jobs from before I committed were clean, but the try on patch set 2
> -
> after the revert have the problem.
>
>
> On 2011/08/09 12:21:40, bsalomon wrote:
>
>> Hmm... having a look. I did run the bots on my windows machine yesterday
>> at
>> 2061 and didn't get any unexpected failures.
>>
>
>  On Mon, Aug 8, 2011 at 8:14 PM, <mailto:vandebo@chromium.org> wrote:
>>
>
>  > The webkit sheriff asked me to revert this change because the WK canary
>> > started
>> > to fail. These tests crashed on Linux: 2d.path.arc.zero.1.html
>> > 2d.path.bezierCurveTo.basic.****html 2d.path.bezierCurveTo.shape.****
>> html
>> > 2d.path.fill.closed.****unaffected.html 2d.path.fill.winding.subtract.*
>> ***
>> > 3.html
>> > 2d.path.moveTo.basic.html 2d.path.rect.end.1.html
>> > 2d.path.rect.selfintersect.****html 2d.path.rect.zero.4.html
>> > 2d.path.stroke.prune.arc.html 2d.path.stroke.prune.rect.html
>> > 2d.path.stroke.union.html 2d.shadow.canvas.basic.html
>> > 2d.shadow.enable.x.html
>> > 2d.state.saverestore.****globalCompositeOperation.html
>> > context.unrecognised.****nullsuffix.html canvas-copyPixels.html
>> > canvas-drawImage-incomplete.****html canvas-fillPath-pattern-****
>> shadow.html
>> > canvas-font-ex-units-crash.****html
>> >
>> >
http://build.chromium.org/p/****chromium.webkit/builders/**<http://build.chro...
>> >
>>
>
> Webkit%20Linux/builds/11653<ht**tp://build.chromium.org/p/**
>
chromium.webkit/builders/**Webkit%20Linux/builds/11653<http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Linux/builds/11653>
> >
>
>> >
>> > Any idea Brian?
>> >
>> >
>> >
>>
>
> http://codereview.chromium.****org/7600001/%3Chttp://coderevi**
> ew.chromium.org/7600001/ <http://codereview.chromium.org/7600001/>>
>
>  >
>>
>
>
>
>
http://codereview.chromium.**org/7600001/<http://codereview.chromium.org/7600...
>

[bsalomon, 2011-08-09 18:03:19]

r2084 has the fix.

On Tue, Aug 9, 2011 at 1:16 PM, Brian Salomon <bsalomon@google.com> wrote:

> I thinkr 2061 introduced a bug and changes in WebKit exposed it. There can
> be a crash at GrContext teardown time. I've reproduced it locally by pulling
> r2061 or skia and TOT WebKit and am working on a fix.
>
>
> On Tue, Aug 9, 2011 at 1:09 PM, <vandebo@chromium.org> wrote:
>
>> I suspect something changed in Webkit (maybe something was reverted),
>> because
>> the try jobs from before I committed were clean, but the try on patch set
>> 2 -
>> after the revert have the problem.
>>
>>
>> On 2011/08/09 12:21:40, bsalomon wrote:
>>
>>> Hmm... having a look. I did run the bots on my windows machine yesterday
>>> at
>>> 2061 and didn't get any unexpected failures.
>>>
>>
>>  On Mon, Aug 8, 2011 at 8:14 PM, <mailto:vandebo@chromium.org> wrote:
>>>
>>
>>  > The webkit sheriff asked me to revert this change because the WK canary
>>> > started
>>> > to fail. These tests crashed on Linux: 2d.path.arc.zero.1.html
>>> > 2d.path.bezierCurveTo.basic.****html 2d.path.bezierCurveTo.shape.****
>>> html
>>> > 2d.path.fill.closed.****unaffected.html 2d.path.fill.winding.subtract.
>>> ****
>>> > 3.html
>>> > 2d.path.moveTo.basic.html 2d.path.rect.end.1.html
>>> > 2d.path.rect.selfintersect.****html 2d.path.rect.zero.4.html
>>> > 2d.path.stroke.prune.arc.html 2d.path.stroke.prune.rect.html
>>> > 2d.path.stroke.union.html 2d.shadow.canvas.basic.html
>>> > 2d.shadow.enable.x.html
>>> > 2d.state.saverestore.****globalCompositeOperation.html
>>> > context.unrecognised.****nullsuffix.html canvas-copyPixels.html
>>> > canvas-drawImage-incomplete.****html canvas-fillPath-pattern-****
>>> shadow.html
>>> > canvas-font-ex-units-crash.****html
>>> >
>>> >
http://build.chromium.org/p/****chromium.webkit/builders/**<http://build.chro...
>>> >
>>>
>>
>> Webkit%20Linux/builds/11653<ht**tp://build.chromium.org/p/**
>>
chromium.webkit/builders/**Webkit%20Linux/builds/11653<http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Linux/builds/11653>
>> >
>>
>>> >
>>> > Any idea Brian?
>>> >
>>> >
>>> >
>>>
>>
>> http://codereview.chromium.****org/7600001/%3Chttp://coderevi**
>> ew.chromium.org/7600001/ <http://codereview.chromium.org/7600001/>>
>>
>>  >
>>>
>>
>>
>>
>>
http://codereview.chromium.**org/7600001/<http://codereview.chromium.org/7600...
>>
>
>

[vandebo (ex-Chrome), 2011-08-09 18:06:22]

epoger, do you want to take it from here?

On 2011/08/09 18:03:19, bsalomon wrote:
> r2084 has the fix.

[epoger, 2011-08-09 18:11:39]

On 2011/08/09 18:06:22, vandebo wrote:
> epoger, do you want to take it from here?
> 
> On 2011/08/09 18:03:19, bsalomon wrote:
> > r2084 has the fix.

Sure, I'm Skia gardener today-Thurs.  I will follow up with Brian.

[epoger, 2011-08-09 19:51:44]

On 2011/08/09 18:11:39, epoger wrote:
> On 2011/08/09 18:06:22, vandebo wrote:
> > epoger, do you want to take it from here?
> > 
> > On 2011/08/09 18:03:19, bsalomon wrote:
> > > r2084 has the fix.
> 
> Sure, I'm Skia gardener today-Thurs.  I will follow up with Brian.

I'm now working on http://codereview.chromium.org/7605006/ ('Skia DEPS roll
2051->2088').  I probably won't be able to commit it until tomorrow morning...
if anyone has an urgent need for a DEPS roll before then, please let me know.