Bug 44643 – Random media test timeouts and crashes on Leopard Commit Bot

• Status: RESOLVED
• Resolution: WONTFIX
• Component: Media

[Eric Seidel (no email)]

Random media test timeouts and crashes on Leopard Commit Bot

Crash logs attached.

[Eric Seidel (no email)]

Created attachment 65485 [details]
crash log #1

[Eric Seidel (no email)]

Created attachment 65486 [details]
Crash log #2

[Eric Seidel (no email)]

Created attachment 65487 [details]
crash log #3

[Eric Seidel (no email)]

Created attachment 65488 [details]
crash log #4

[Eric Seidel (no email)]

Created attachment 65489 [details]
Crash log #5

[Eric Seidel (no email)]

I have 15 more crash logs on the machine as well.  I'm able to correlate the individual logs with tests by looking at the logs if that would be helpful.

[Eric Seidel (no email)]

Here are recent tests to crash for the commit-queue:

media/video-aspect-ratio.html -> crashed
media/video-layer-crash.html -> crashed
media/video-aspect-ratio.html -> crashed
media/video-preload.html -> crashed
media/video-preload.html -> crashed
media/video-preload.html -> crashed
media/video-preload.html -> crashed
media/video-zoom.html -> crashed
media/video-preload.html -> crashed
media/video-preload.html -> crashed
media/video-aspect-ratio.html -> crashed
media/video-aspect-ratio.html -> crashed
media/video-preload.html -> crashed
media/video-preload.html -> crashed
media/video-preload.html -> crashed
media/event-attributes.html -> crashed
media/video-load-readyState.html -> crashed
media/video-preload.html -> crashed

[Eric Seidel (no email)]

Note, the crashes above are in chronological order (oldest first) and are *all* the crashes.  So it's actually possible to correlate these w/o looking more specifically at the times in the logs (although I'm happy to do so if you like) because crash-log #1 is the most recent crash from the bot, and the last test in comment #7 is the most recent to crash on the bot. :)

[Eric Seidel (no email)]

By *all* the crashes in comment 8, I was attempting to imply that I hadn't filtered out non-media crashes.  Comment 8 is a list of the 18 most recent tests to crash, according to the commit-queue logs.  There are older crashes of course. :)

[Eric Seidel (no email)]

fast/forms/search-event-delay.html -> failed
media/controls-drag-timebar.html -> failed
fast/css/font-face-svg-decoding-error.html -> failed
fast/css/font-face-svg-decoding-error.html -> failed
fast/css/font-face-svg-decoding-error.html -> failed
fast/loader/recursive-before-unload-crash.html -> failed
fast/loader/recursive-before-unload-crash.html -> failed
fast/loader/recursive-before-unload-crash.html -> failed
fast/loader/recursive-before-unload-crash.html -> failed
fast/loader/recursive-before-unload-crash.html -> failed
fast/loader/recursive-before-unload-crash.html -> failed
animations/combo-transform-translate+scale.html -> failed
animations/combo-transform-translate+scale.html -> failed
media/video-played-collapse.html -> failed
media/video-played-collapse.html -> failed

Since I also mentioned timeouts in the title of this bug, I should also tell you what  the 15 most recent non-crash failures on the CQ were.  Note this list is also media-heavy, due to media-related timeouts.  (For this list I did filter out a flaky XSS auditor test which was failing constantly yesterday and has since been fixed.)

[Simon Fraser (smfr)]

Thanks for the logs. All over the place!

[Eric Seidel (no email)]

I've uninstalled Flip4Mac on that machine in case that helps (it came by default since it's a Google image).  It's also possible it somehow has mis-matched quicktime libraries, but I doubt it.  We'd be able to tell from the CrashLogs by checking the library hashes against known good ones.

I'll email you all 20 logs on the machine in case those are helpful.

[Eric Seidel (no email)]

Btw, these are the last of the crash reports we should get as I'm about to skip the media tests on the commit-queue as part of bug 38912.  If we need more, I can always unskip it.

[Brent Fulgham]

Whatever this was, we aren't going to worry about it now that more than a decade has passed.