Bug 69240 – [Chromium] 10% intl1 PLT regression Sept 28

• Component: Page Loading
• Status: RESOLVED FIXED

[Adam Barth]

http://build.chromium.org/f/chromium/perf/xp-release-dual-core/intl1/report.html?history=130&rev=103677

This appears to be caused by this Skia roll r2351:r2358:
http://src.chromium.org/viewvc/chrome?view=rev&revision=103204

[Adam Barth]

http://code.google.com/p/skia/source/list?num=7&start=2358

[Adam Barth]

Maybe this change http://code.google.com/p/skia/source/detail?r=2358 ?

[Mike Reed]

This is in process of being fixed/reversed.

[Adam Barth]

> This is in process of being fixed/reversed.

I'm concerned that Bug 67876 has been unresolved for upwards of a month.  It's important that we not carry these regression in the tree because they mask other regressions.  Do you have an estimate for when this issue will be resolved?

[Mike Reed]

Should be fixed in latest skia DEPS roll.

[Adam Barth]

> Should be fixed in latest skia DEPS roll.

Thanks Mike.  It looks on this graph that we've gotten most (if not all) of the performance back:

http://build.chromium.org/f/chromium/perf/xp-release-dual-core/intl1/report.html?history=150&rev=103710

This improvement appears to have been cause by rolling Skia:

http://src.chromium.org/viewvc/chrome?view=rev&revision=103704

Just to confirm before closing this bug: That improvement is the fix for the Sept 28 regression and not for the regression we're tracking in Bug 67876 ?  (They are of a similar magnitude, which makes it hard for me to tell just from the graph.)

[Mike Reed]

I think so. The recent regression was definitely due to 2358, which was (effectively) reverted.

[Adam Barth]

Thanks for the quick response.