Roll deps for Skia r626:632.

Roll deps for Skia r626:632.

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=68558

[Lei Zhang, 2010-12-07 21:21:16]

Per Mike's request.

[Stephen White, 2010-12-07 21:33:11]

On 2010/12/07 21:21:16, Lei Zhang wrote:
> Per Mike's request.

Have you run this by the layout test trybots?  In particular, I'm wondering if
r630 will affect any pixel results (631 and 632 should be ok though).

[Lei Zhang, 2010-12-07 21:35:33]

On 2010/12/07 21:33:11, Stephen White wrote:
> On 2010/12/07 21:21:16, Lei Zhang wrote:
> > Per Mike's request.
> 
> Have you run this by the layout test trybots?  In particular, I'm wondering if
> r630 will affect any pixel results (631 and 632 should be ok though).

No, I'll do that now.

[Lei Zhang, 2010-12-07 22:57:39]

On 2010/12/07 21:35:33, Lei Zhang wrote:
> On 2010/12/07 21:33:11, Stephen White wrote:
> > On 2010/12/07 21:21:16, Lei Zhang wrote:
> > > Per Mike's request.
> > 
> > Have you run this by the layout test trybots?  In particular, I'm wondering
if
> > r630 will affect any pixel results (631 and 632 should be ok though).
> 
> No, I'll do that now.

It's been a while since the last time I dealt with layout tests. I got the
results back from the linux layout bot. How do I tell which ones are genuine
failures and which are just flakiness? Do I need to update the layout tests in
this CL?

[Stephen White, 2010-12-07 23:15:35]

On 2010/12/07 22:57:39, Lei Zhang wrote:
> On 2010/12/07 21:35:33, Lei Zhang wrote:
> > On 2010/12/07 21:33:11, Stephen White wrote:
> > > On 2010/12/07 21:21:16, Lei Zhang wrote:
> > > > Per Mike's request.
> > > 
> > > Have you run this by the layout test trybots?  In particular, I'm
wondering
> if
> > > r630 will affect any pixel results (631 and 632 should be ok though).
> > 
> > No, I'll do that now.
> 
> It's been a while since the last time I dealt with layout tests. I got the
> results back from the linux layout bot. How do I tell which ones are genuine
> failures and which are just flakiness? Do I need to update the layout tests in
> this CL?

In theory, they should be green (assuming the main bots are green).  They're
usually pretty reliable.

One way to check your results would be to send a whitespace change to the
trybots and see if you get the same fails.  (Sorry, I know this is a pain, but
it's better than getting reverted if it makes the main waterfall go red).

I haven't actually rolled a breaking change from Skia since the test results
were upstreamed to WebKit, and I suspect this is actually a bit of a pain right
now.  The safest thing to do is probably to roll Skia DEPS in Chrome, and
temporarily add the failing tests to
webkit/tools/layout_tests/test_expectations.txt in the same CL.

The rebaselining has to be done upstream in WebKit, since that's where the test
results live now.   You might also want to check with agl and morrita (assuming
that is the offending change) to see if they plan to rebaseline.

[Lei Zhang, 2010-12-08 04:12:20]

Since today is a pretty quiet day, I'm going to just roll and then clean up the
failures. Hope that's ok.