[regression] Specific YouTube playback becomes solid (gen6)
Categories
(Core :: Graphics: WebRender, defect, P2)
Tracking
()
Tracking | Status | |
---|---|---|
firefox-esr68 | --- | unaffected |
firefox69 | --- | unaffected |
firefox70 | --- | wontfix |
firefox71 | --- | wontfix |
firefox72 | --- | wontfix |
firefox75 | --- | wontfix |
firefox76 | --- | wontfix |
firefox77 | --- | wontfix |
firefox78 | --- | fixed |
People
(Reporter: yfdyh000, Assigned: kvark)
References
(Blocks 1 open bug, Regression)
Details
(Keywords: regression)
Attachments
(2 files, 1 obsolete file)
This problem occurs in an Sandy Bridge GT1 (HD 2000) device with gfx.webrender.enabled = true. I know the GPU is not officially supported from WR at now, but it's worse than before the results.
STR:
- open Firefox, set
gfx.webrender.enabled
to true in about:config, restart Firefox. - open https://www.youtube.com/watch?v=ZJa7ZF_2H2I and play.
Actual results:
The screen is blank or solid color with the content changes, no character.
In the newer Nightly, the entire window is white flashes, delay, and even Windows to block the Firefox access to the graphics hardware, but it may be another issue.
Expected results:
Normal playback.
Regression range:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=b0a5b50962e7774af1c0500098a7a14170397961&tochange=3bcc094e0863013bd86996c7665260594d6413e1
Bug 1579227 - Switch YUV images to be opaque in WR r=jrmuizel
Assignee | ||
Comment 3•5 years ago
|
||
Does look related, but not expected to cause the issue. I'll take a look!
Assignee | ||
Comment 4•5 years ago
|
||
Is it happening on specific videos for you? What makes them different from the others?
Other YouTube videos playback properly, I only see this video immediately result in a solid color.
Hiding the screenshot, the green may be due to another issue (screen white-flashing, graphics access blocked, in more newer Nightly only).
Comment 6•5 years ago
|
||
:kvark, since this bug is a regression, could you fill (if possible) the regressed_by field?
For more information, please visit auto_nag documentation.
Assignee | ||
Comment 7•5 years ago
|
||
YF, could you try the build from this push - https://treeherder.mozilla.org/#/jobs?repo=try&revision=ec17cfe779b37a0e32d613288f3acaa6e1c0b4e8 please? I manually reverted the suspected change. It's quite strange that it would affect only specific videos and only on specific hardware. It looks more like an issue with image/video decoding to me (cc :aosmond).
The build playing the video correctly.
Assignee | ||
Comment 9•5 years ago
|
||
YF, thank you for testing! I assume you verified along the way in "about:support" that WebRender is used (and enable it if needed).
Jessie, it sounds like we'll need a machine in Toronto to reproduce this.
Updated•5 years ago
|
Comment 11•5 years ago
|
||
That's more of a question for the media team than me. My understanding is that YUV images with transparency are rare, but not impossible, see Y41T and Y42T:
https://docs.microsoft.com/en-us/windows/win32/medfound/recommended-8-bit-yuv-formats-for-video-rendering
However I don't think the media framework spits out partial frames, so it is probably just an internal detail. Are we able to get a local copy of the video to make sure this isn't YouTube serving different people different encodings?
Reporter | ||
Comment 12•5 years ago
|
||
(In reply to Andrew Osmond [:aosmond] from comment #11)
However I don't think the media framework spits out partial frames, so it is probably just an internal detail. Are we able to get a local copy of the video to make sure this isn't YouTube serving different people different encodings?
Use the https://addons.mozilla.org/firefox/addon/youtube_downloader_webx/ 's Download Panel for https://www.youtube.com/watch?v=ZJa7ZF_2H2I, the WEBM 360p can reproduce the problem, MP4 360P cannot reproduce (playback correctly).
Video sample: https://send.firefox.com/download/295073131f81ca74/#dGQt7WhWDlVl695_-LLvCw
Updated•5 years ago
|
Updated•5 years ago
|
Comment 13•5 years ago
|
||
Are you still working on this, Dzmitry? The regression triage meeting was wondering about status flags for various upcoming releases.
Assignee | ||
Comment 14•5 years ago
|
||
Andrew, I stopped at stage where I need the actual device to reproduce. Consider me still working on it, unless someone else wants to pick it up.
Updated•5 years ago
|
Comment 15•5 years ago
|
||
https://bugzilla.mozilla.org/show_bug.cgi?id=1585782#c10
Would that machine do the trick or do we need something else?
Assignee | ||
Comment 16•5 years ago
|
||
The Vostro machine here has G45/G43 chipset and it's unable to even switch WR on because of an Angle blocklist entry established in bug 1153381. Unless you meant a different Vostro machine.
Updated•4 years ago
|
Comment 17•4 years ago
|
||
YF does it work better if you disable picture caching by setting gfx.webrender.picture-caching=false?
Updated•4 years ago
|
Reporter | ||
Comment 18•4 years ago
|
||
(In reply to Jeff Muizelaar [:jrmuizel] from comment #17)
YF does it work better if you disable picture caching by setting gfx.webrender.picture-caching=false?
Due to bug 1586040, Firefox is almost completely unavailable if WR is on at now, so I failed to test it.
Comment 20•4 years ago
|
||
Actually the fix that I think will help got backed out so it's probably best to wait until it relands.
Reporter | ||
Comment 21•4 years ago
|
||
This problem is not reproducible now (bug 1633628 landed), gfx.webrender.picture-caching=true.
Updated•4 years ago
|
Updated•4 years ago
|
Updated•4 years ago
|
Updated•4 years ago
|
Comment 22•4 years ago
|
||
After double checking, clearing QE+ flag since we do not have the specified hardware.
Description
•