Open Bug 1522739 Opened 5 years ago Updated 2 years ago

Intermittent dom/media/test/test_mediarecorder_pause_resume_video.html | Should not have any resize events!

Categories

(Core :: Audio/Video: Recording, defect, P3)

defect

Tracking

()

People

(Reporter: intermittent-bug-filer, Unassigned)

References

(Regression)

Details

(Keywords: intermittent-failure, leave-open, regression, Whiteboard: [stockwell disabled])

Attachments

(4 files)

Filed by: cbrindusan [at] mozilla.com

https://treeherder.mozilla.org/logviewer.html#?job_id=223915288&repo=mozilla-inbound

https://queue.taskcluster.net/v1/task/Vk9UHw2tT3K-DjbXqTBksQ/runs/0/artifacts/public/logs/live_backing.log

00:24:15 INFO - TEST-START | dom/media/test/test_mediarecorder_pause_resume_video.html
00:25:04 INFO - TEST-INFO | started process screenshot
00:25:12 INFO - TEST-INFO | screenshot: exit 0
00:25:12 INFO - Buffered messages logged at 00:24:15
00:25:12 INFO - Drawing color 255,0,0,1
00:25:12 INFO - TEST-PASS | dom/media/test/test_mediarecorder_pause_resume_video.html | Media recorder stream = canvas stream at the start of recording
00:25:12 INFO - TEST-PASS | dom/media/test/test_mediarecorder_pause_resume_video.html | Media recorder should be recording
00:25:12 INFO - Buffered messages logged at 00:24:24
00:25:12 INFO - Got 'start' event
00:25:12 INFO - TEST-PASS | dom/media/test/test_mediarecorder_pause_resume_video.html | Media recorder should be paused
00:25:12 INFO - Drawing color 0,0,255,1
00:25:12 INFO - Buffered messages logged at 00:24:50
00:25:12 INFO - Drawing color 0,255,0,1
00:25:12 INFO - Buffered messages logged at 00:24:55
00:25:12 INFO - Got 'dataavailable' event
00:25:12 INFO - Got 'stop' event
00:25:12 INFO - TEST-PASS | dom/media/test/test_mediarecorder_pause_resume_video.html | Media recorder should be incative after stop
00:25:12 INFO - TEST-PASS | dom/media/test/test_mediarecorder_pause_resume_video.html | Expected 1 dataavailable event
00:25:12 INFO - TEST-PASS | dom/media/test/test_mediarecorder_pause_resume_video.html | Should have gotten a data blob
00:25:12 INFO - TEST-PASS | dom/media/test/test_mediarecorder_pause_resume_video.html | video element should be same size as canvas once metadata is loaded
00:25:12 INFO - Buffered messages finished
00:25:12 INFO - TEST-UNEXPECTED-FAIL | dom/media/test/test_mediarecorder_pause_resume_video.html | Should not have any resize events!

Status: NEW → RESOLVED
Closed: 5 years ago
Resolution: --- → INCOMPLETE
Status: RESOLVED → REOPENED
Resolution: INCOMPLETE → ---
Assignee: nobody → apehrson
Status: REOPENED → ASSIGNED
Flags: needinfo?(apehrson)
Priority: P5 → P3
Component: Audio/Video: Playback → Audio/Video: Recording

The reason this is spiking now, is because per spec we're now queueing a task to stop gathering data into the blob, rather than doing it directly. Bug 1514158 added that.

I'll linearize and fix the test to account for this, and try to see if there's still a low rate of failures occuring (as it were prior to bug 1514158).

Has Regression Range: --- → yes
Keywords: regression
Regressed by: 1514158
Attachment #9102567 - Attachment description: Bug 1522739 - Disable test_mediarecorder_pause_resume_video.html on win&linux debug and asan. → Bug 1522739 - Disable test_mediarecorder_pause_resume_video.html on win&linux debug and asan. r?#intermittent-reviewers

Jan, could you please review the patches from Andreas to see if it fixes these really frequent failures? Thank you.

Flags: needinfo?(jib)
Attachment #9102567 - Attachment description: Bug 1522739 - Disable test_mediarecorder_pause_resume_video.html on win&linux debug and asan. r?#intermittent-reviewers → Bug 1522739 - Disable test_mediarecorder_pause_resume_video.html on linux-qr debug and win10 debug and asan. r?#intermittent-reviewers
Pushed by csabou@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/8ff2406c2124
Disable test_mediarecorder_pause_resume_video.html on linux-qr debug and win10 debug and asan. r=gbrown
Keywords: leave-open
Whiteboard: [stockwell disable-recommended] → [stockwell disabled]

Pehrsons and I discussed offline about taking a different path on this, perhaps backup up to earlier behavior on some of this. He said he'd work on putting up new patches tomorrow.

Flags: needinfo?(jib) → needinfo?(apehrson)
Depends on: 1590319

This should occur less again, after bug 1586370 relaxed the timing constraints in this test a bit (it made it spike). The patches I were working on still made this fail intermittently in test-verify, and they never got ready to land. We set the timestamp of a frame to the vsync time that triggered the requestAnimationFrame callback. But what we're testing in the test is the order of pause/resume and the requestAnimationFrame callback occurring.

This could be solved by using one timestamp for the capture time, i.e., when the requestAnimationFrame callback ran, for ordering with pause/resume, and another for rendering time, i.e., how to sync it to an audio track.

I won't be able to work on this anytime soon, so unassigning.

Assignee: apehrson → nobody
Status: ASSIGNED → NEW
Flags: needinfo?(apehrson)

This is now occurring on TV and mda tests.

Nils, based on the above comment, can you assign someone to take a look?

Flags: needinfo?(drno)

Based on IRC convo, please disregard the above:

dminor: apavel|sheriffduty: I think https://treeherder.mozilla.org/#/jobs?repo=autoland&fromchange=d01f200c7b3ac2a864362d6c01bbbe4e50962301&selectedJob=281948363 is fallout from bug 1586370 which my patch was an attempt to fix.

dminor: I'm going to prepare a patch to disable that test, we have bug 1605134 on file to follow up on it anyway.

Flags: needinfo?(drno)

Dan, this is failing frequently also on linux64 and windows7-32: https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?startday=2019-12-13&endday=2019-12-20&tree=trunk&bug=1522739
Should we consider disabling further or is there a fix in the works?

Flags: needinfo?(dminor)

I made a mistake when I tried the fix that Andreas suggested, so I'm going to redo that and see if it helps. If it is still intermittent, I'll disable it on all platforms.

Flags: needinfo?(dminor)

I'd like to see what happens with this try job before disabling the test completely:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=2d9d6a15d4d231d1c34ba567b427a83bf3642bd0

Dan, the patch seems to be working https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=282155687&repo=try&lineNumber=2823 do you mind making a patch so we can get this landed on tree and avoid the high number of failures? Thank you.

Flags: needinfo?(dminor)

Patch is up for review at Bug 1605703.

Flags: needinfo?(dminor)

No failures since the patch was merge into central on the 24th.
Paul, Bryce (as https://hg.mozilla.org/mozilla-central/rev/3e577500606bd93a80b217d5c7765233d5057995 points to you as being the test owner) should this be un-skipped now that Bug 1605703 it is fixed?
Currently test is skipped partly because of this bug on:

skip-if = toolkit == 'android' || (webrender && debug && os == "linux") || (os == "win" && os_version == "10.0" && (debug || asan)) # android(bug 1232305) win&linux bug 1522739

Flags: needinfo?(padenot)
Flags: needinfo?(bvandyk)

Joel is the above something you can answer?

Flags: needinfo?(jmaher)

the only way to determine that is to push it to try with it not skipped and retrigger it 100 times.

Flags: needinfo?(jmaher)

(In reply to Joel Maher ( :jmaher ) (UTC-4) from comment #46)

the only way to determine that is to push it to try with it not skipped and retrigger it 100 times.

Thank you.

Try push: https://treeherder.mozilla.org/#/jobs?repo=try&revision=f39a40a501a94ca96e2087882294142e697aa636

Will update here once the jobs are done.

Flags: needinfo?(padenot)
Flags: needinfo?(bvandyk)

Based on try results, i believe this can be unskipped. Attaching patch.

Assignee: nobody → apavel
Status: NEW → ASSIGNED

If the test is still skipped on android, do we leave [stockwell disabled]? or leave-open in the keyword?

Pushed by apavel@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/d8a2c33a4c7a
unskip test_mediarecorder_pause_resume_video.html r=jmaher

leave [stockwell disabled] in the whiteboard tag

Understood, thank you for clarifying.

Assignee: apavel → nobody
Status: ASSIGNED → NEW
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.