roll skia to 4332

roll skia to 4332
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=144160

[reed1, 2012-06-25 13:09:55]

control #5
https://chromiumcodereview.appspot.com/10533136

[reed1, 2012-06-25 15:24:18]

only failure I quested was 2d.fillText.gradient.html on debug-linux, but it
draws fine for me locally, and the other bots don't mention it, so I'm going to
stick my head in the sand.

Waiting on windows... (big shock)

[reed1, 2012-06-25 16:13:13]

correction: 2d.fillText.gradient.html fails on all DEBUG bots. Investigating...

[Tom Hudson, 2012-06-25 17:29:11]

Nothing else looks bad on the win bots, now that they've finished.

[reed1, 2012-06-25 18:35:32]

failure is in gpu/textcontext. handing over to that team.

[Tom Hudson, 2012-06-25 18:42:07]

On 2012/06/25 18:35:32, reed1 wrote:
> failure is in gpu/textcontext. handing over to that team.

The cause is r4313, qv http://codereview.appspot.com/6324046/. We'd like to have
setPaint() called only once, and there is now an assert to enforce it.

Both GrBatchedTextContext::init() and GrDefaultTextContext::init() were causing
setPaint() to be called, but that was fixed in the initial commit. Presumably
Chromium is hitting a different bad codepath.

[reed1, 2012-06-25 20:27:05]

up the rev. to 4332, which includes a work-around for the textcontext issue.