webkit_unit_tests failing on chromium.linux android bots

Isaac Levy

unread,

Mar 17, 2013, 1:27:23 AM3/17/13

to webkit-gardening, John Knottenbelt

Android testers on chromium.webkit and chromium.linux are failing with high probability on: webkit_unit_test::CompositorScrollIsUserScrollShortPage

and a couple others.

Cause is http://wkrev.com/145898, chromium issues tracked by crbug.com/196687

It appears the roll occurred while the android chromium.webkit tester was still trying the revisions (though after the step turned red). Is the gardening tool ignoring android?

Could next gardener coordinate w/ sheriff and handle? Not sure what procedure is (rollback CL vs. disabling tests). Android tests can be disabled here:

http://src.chromium.org/viewvc/chrome/trunk/src/build/android/pylib/gtest/filter/webkit_unit_tests_disabled?view=markup

Thanks,

Isaac

Isaac Levy

unread,

Mar 17, 2013, 1:29:14 AM3/17/13

to Isaac Levy, Jochen Eisinger, webkit-gardening, John Knottenbelt

Jochen helped rollout the CL -- thanks!

Isaac Levy

unread,

Mar 17, 2013, 1:43:55 AM3/17/13

to Isaac Levy, Jochen Eisinger, webkit-gardening, John Knottenbelt

(note that chromium.linux will continue to fail until DEPS are rolled, so this still needs action by gardener)

Pavel Feldman

unread,

Mar 19, 2013, 12:26:45 AM3/19/13

to Isaac Levy, John Knottenbelt, webkit-g...@chromium.org, Jochen Eisinger

I did not use CQ during my shift - it was busted during the duty and infrastructure team knows it. My PST counterparts barely rolled anything due to that reason, so I was landing by hand based on chromium.webkit results. I don't think I have android bots in my filtered gardening bot set, I'll add them for my next duty.

The roll in question is a follow up to the previous one that resulted is perf/size failure. Such things happen and we are lucky I was rolling often - the reason was easily identified. You either roll out such a WebKit roll or you fix thing upstream live and roll again. Given that it was only a handful of revisions and none seemed major change, and it was 6am PST (low traffic) I took the risk.

Now my recollection is that main waterfall was green after that roll, no sheriff pinged me on the channel, the tree remained green. So I spent the rest of the evening fixing the Perf bot compile error that was introduced hours earlier.

I am sorry if this caused you trouble, but the rolls must happen and must happen often. This failure looks minor and manageable given tree is watched comparing to the potential perf / other regressions caused by upstream.

Regards
Pavel

Reply all

Reply to author

Forward