Could just be coincidence, but it seems odd that this failed with the same number of events on both mozilla-central and TraceMonkey during the same hour, the hour when everything's busy with nightlies and l10n repacks of them. http://tinderbox.mozilla.org/showlog.cgi?log=TraceMonkey/1297078413.1297082164.4830.gz Rev3 MacOSX Snow Leopard 10.6.2 tracemonkey debug test mochitests-4/5 on 2011/02/07 03:33:33 s: talos-r3-snow-026 1006 INFO TEST-PASS | /tests/layout/base/tests/test_bug629838.html | Background color iteration 46, afterpaint count: 25, mozpaint count: 55 1007 INFO TEST-PASS | /tests/layout/base/tests/test_bug629838.html | mozPaintCount incremented enough from background color changes. 1008 ERROR TEST-UNEXPECTED-FAIL | /tests/layout/base/tests/test_bug629838.html | Got MozAfterPaint event 25 time(s). http://tinderbox.mozilla.org/showlog.cgi?log=TraceMonkey/1297078413.1297082164.4830.gz Rev3 MacOSX Snow Leopard 10.6.2 tracemonkey debug test mochitests-4/5 on 2011/02/07 03:33:33 s: talos-r3-snow-026 1006 INFO TEST-PASS | /tests/layout/base/tests/test_bug629838.html | Background color iteration 46, afterpaint count: 25, mozpaint count: 55 1007 INFO TEST-PASS | /tests/layout/base/tests/test_bug629838.html | mozPaintCount incremented enough from background color changes. 1008 ERROR TEST-UNEXPECTED-FAIL | /tests/layout/base/tests/test_bug629838.html | Got MozAfterPaint event 25 time(s).

I suspected that

Doh, pressed enter by accident. Anyway, I suspected that this test might be vulnerable to machine load. I could try and fix it by adjusting the constants, but I think it's probably just better to fall back on merely testing that we get enough afterpaint events 'eventually', and drop the strict timing requirements since they seem to be unreliable.

I've attached a patch that changes the assertions in this test so that it merely requires MozAfterPaint to fire a certain number of times eventually, instead of asserting that it fires in sync with mozPaintCount increases. This should be reliable, even if it lets later breakages through.

Comment on attachment 512595 [details] [diff] [review] No longer assert that MozAfterPaint fires in sync with mozPaintCount increases. Kevin, you should ask someone to review the patch.

Comment on attachment 512595 [details] [diff] [review] No longer assert that MozAfterPaint fires in sync with mozPaintCount increases. I think roc can review this.

Comment on attachment 512595 [details] [diff] [review] No longer assert that MozAfterPaint fires in sync with mozPaintCount increases. Review of attachment 512595 [details] [diff] [review]:

Thanks Kevin for your patch. Guys, can one of you please land this?

Pushed: https://bugzilla.mozilla.org/show_bug.cgi?id=632102

http://hg.mozilla.org/projects/cedar/rev/28b52c3fd694 http://hg.mozilla.org/mozilla-central/rev/28b52c3fd694