REOPENED 221693
[GPUP] One media test fails or timeout when media in GPU Process is enabled 
https://bugs.webkit.org/show_bug.cgi?id=221693
Summary [GPUP] One media test fails or timeout when media in GPU Process is enabled
Peng Liu
Reported 2021-02-10 10:29:07 PST
media/audio-playback-restriction-removed-track-enabled.html [ Timeout Pass ] media/controls/inline-elements-dropoff-order.html [ Failure Pass ] media/encrypted-media/clearKey/clearKey-session-life-cycle.html [ Failure Crash ] media/media-can-play-webm.html [ Failure ] media/media-controller-time-constant.html [ Timeout ] media/media-controller-time.html [ Timeout ] media/media-controller-unpause.html [ Timeout ] media/track/track-cues-pause-on-exit.html [ Timeout Pass ] media/video-as-img-output-pts.html [ Timeout Pass ] media/video-canvas-createPattern.html [ Crash ] media/video-controller-child-rate.html [ Timeout ] media/video-controller-currentTime-rate.html [ Failure ] media/video-orientation-canvas.html [ Failure ] media/video-set-rate-from-pause.html [ Timeout Pass ]
Attachments
Patch (1.77 KB, patch)
2021-03-04 10:41 PST, Peng Liu 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Philippe Normand
Comment 1 2021-02-17 08:31:34 PST
Why was this flagged in the wk2 expectations?
Peng Liu
Comment 2 2021-02-17 09:47:25 PST
They are flagged as flaky because of the GPU Process work. Now "media in GPU Process" is enabled in the test runner for Mac and iOS. It is disabled for GStreamer-based ports, right? Then maybe I need to move the change to mac-wk2 and ios-wk2.
Philippe Normand
Comment 3 2021-02-17 09:54:09 PST
(In reply to Peng Liu from comment #2) > They are flagged as flaky because of the GPU Process work. Now "media in GPU > Process" is enabled in the test runner for Mac and iOS. It is disabled for > GStreamer-based ports, right? Then maybe I need to move the change to > mac-wk2 and ios-wk2. I moved most to the glib TestExpectation as expect-pass: https://github.com/WebKit/WebKit/commit/3ff6c67518b693e0511fa75b1342fdcdd1b24941
Peng Liu
Comment 4 2021-02-17 10:00:23 PST
(In reply to Philippe Normand from comment #3) > > I moved most to the glib TestExpectation as expect-pass: > https://github.com/WebKit/WebKit/commit/ > 3ff6c67518b693e0511fa75b1342fdcdd1b24941 Thanks!
Radar WebKit Bug Importer
Comment 5 2021-02-17 10:30:30 PST
<rdar://problem/74439562>
Amir Mark Jr
Comment 6 2021-03-02 12:50:58 PST
media/video-controller-child-rate.html is consistently failing on iOS. Able to reproduce issue using: run-webkit-tests --ios-simulator --iterations 200 -f media/video-controller-child-rate.html
Peng Liu
Comment 7 2021-03-03 15:03:01 PST
Following tests are passing on bots now: media/video-as-img-output-pts.html media/track/track-cues-pause-on-exit.html media/controls/inline-elements-dropoff-order.html media/audio-playback-restriction-removed-track-enabled.html
Peng Liu
Comment 8 2021-03-04 10:41:45 PST
Created attachment 422256 [details] Patch
EWS
Comment 9 2021-03-04 13:00:57 PST
Committed r273914: <https://commits.webkit.org/r273914> All reviewed patches have been landed. Closing bug and clearing flags on attachment 422256 [details].
Peng Liu
Comment 10 2021-03-04 13:05:11 PST
Reopen this bug to track the remaining 4 test failures.
Jon Lee
Comment 11 2021-03-05 01:36:45 PST
media/media-controller-unpause.html no longer fails on the bots, likely due to r273895
Eric Carlson
Comment 12 2021-03-10 13:25:17 PST
It looks like these are left: media/encrypted-media/clearKey/clearKey-session-life-cycle.html media/video-controller-child-rate.html media/video-set-rate-from-pause.html
Eric Carlson
Comment 13 2021-03-17 14:20:07 PDT
media/video-set-rate-from-pause.html was fixed by https://bugs.webkit.org/show_bug.cgi?id=223386
Peng Liu
Comment 14 2021-03-19 10:25:32 PDT
Patch for Bug 223523 will fix the failure of media/video-controller-child-rate.html.
Jer Noble
Comment 15 2021-03-21 14:31:20 PDT
media/encrypted-media/clearKey/clearKey-session-life-cycle.html is failing because the retain count of the mediaKeySession internal instance is 2 instead of 1, but this is due to the instance being retained by the RemoteCDMFactory for tracking purposes, and not an actual bug.
Note You need to log in before you can comment on or make changes to this bug.