Video has visible artifacts and distortions for the brief time after switching from another tab after landing patch from bug #1758605
Categories
(Core :: Audio/Video: Playback, defect, P3)
Tracking
()
Tracking | Status | |
---|---|---|
firefox-esr91 | --- | unaffected |
firefox98 | --- | unaffected |
firefox99 | --- | unaffected |
firefox100 | --- | fixed |
firefox101 | --- | fixed |
People
(Reporter: Virtual, Unassigned)
Details
(Keywords: nightly-community, regression, Whiteboard: [nightly-community])
Attachments
(4 files)
Video has visible artifacts and distortions for the brief time after switching from another tab after landing patch from bug #1758605. It is also worth to notice that when you switch to video, it starts to play with delay.
Regression range pushlog:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=1fd694e26ee5eaefed65f75974bd6e8a5001a1cf&tochange=88a481adf83070d1cdcc2bb074f6545b5dfe97e7
mozillaregression-GUI points that regression is caused by:
Narrowed integration regression window from [2fe7954f, 88a481ad] (3 builds) to [1fd694e2, 88a481ad] (2 builds) (~1 steps left)
Found commit message:
Bug 1758605 - Fix MediaCapabilities not detecting hardware support for certain H264 profiles. r=alwu
Added two missing flags to H264::CreateExtraData that are required to parse the avcC box correctly.
Differential Revision:
https://phabricator.services.mozilla.com/D140695
Steps to reproduce:
- Open latest Mozilla Firefox Nighty 100.0a1 (2022-04-01) (64-bit)
- Open https://coub.com/view/2htg0f
- Play clip
- Open new tab
- Get back to tab with clip
- Redo step 4 and 5 few times
- Notice that video has visible artifacts and distortions for the brief time after switching from another tab
Observed results:
Video has visible artifacts and distortions for the brief time after switching from another tab.
Expected results:
Video does not have visible artifacts and distortions.
Reporter | ||
Updated•3 years ago
|
Hmm, I can't reproduce the distortion on my machine, although I am running a much newer Nvidia card. However, the issue may be caused by slower initialization of the decoder after the hardware AV1 decoding patch in that pushlog. If so, that should be fixed in the patch for Bug 1760804, once that lands. That should also reduce the delay before you see video playback resume after the decoder is suspended when switching tabs.
Let me know if the issue persists after that patch lands, hopefully that is the issue.
Out of curiosity, though, could you upload your DxDiag here as well? I'm curious what Windows Media Foundation decoders you have available on your system. It's possible that the change actually made your Firefox use an Nvidia-provided hardware H264 decoder, although it seems unlikely from my experience.
Reporter | ||
Comment 5•3 years ago
|
||
(In reply to Zaggy1024 from comment #4)
Out of curiosity, though, could you upload your DxDiag here as well? I'm curious what Windows Media Foundation decoders you have available on your system. It's possible that the change actually made your Firefox use an Nvidia-provided hardware H264 decoder, although it seems unlikely from my experience.
Sure, please have a look.
Odd, I guess Windows 7's version of DxDiag show the Media Foundation decoders. That's all right, though, we can see if the incoming patch fixes the issue. It should be in autoland today.
Reporter | ||
Comment 7•3 years ago
|
||
(In reply to Zaggy1024 from comment #4)
Hmm, I can't reproduce the distortion on my machine, although I am running a much newer Nvidia card. However, the issue may be caused by slower initialization of the decoder after the hardware AV1 decoding patch in that pushlog. If so, that should be fixed in the patch for Bug 1760804, once that lands. That should also reduce the delay before you see video playback resume after the decoder is suspended when switching tabs.
Let me know if the issue persists after that patch lands, hopefully that is the issue.
Unfortunately the issue persists in the latest Mozilla Firefox Nightly 100.0a1 (2022-04-03) (64-bit) [Built from https://hg.mozilla.org/mozilla-central/rev/b9165a6769deef9d17c8a472b8396f38cad2f894]
Would you be able to test on this revision? That's where the revision landed to fix decoder init. The build you linked was after the revision was backed out for another regression, there is a fix incoming for that as well.
Updated•3 years ago
|
Updated•3 years ago
|
Comment 9•3 years ago
|
||
Set release status flags based on info from the regressing bug 1758605
Reporter | ||
Comment 10•3 years ago
|
||
(In reply to Zaggy1024 from comment #8)
Would you be able to test on this revision? That's where the revision landed to fix decoder init. The build you linked was after the revision was backed out for another regression, there is a fix incoming for that as well.
I'm not seeing any test builds with this revision, but the issue persists in the latest Mozilla Firefox Nightly 101.0a1 (2022-04-06) (64-bit) [Built from https://hg.mozilla.org/mozilla-central/rev/cd1e11d59e128a09f246adfb9211a1d894d623bc].
Comment 11•3 years ago
|
||
In the bug description you mentioned "It is also worth to notice that when you switch to video, it starts to play with delay." Has the delay reduced in the current Nightly build? The patch that improved WMF decoder init time should have landed that revision, so I'm curious whether it's improved that aspect of the bug.
It may be very tricky to nail down this bug, though, since most of the behavior related to H264 should essentially the same as before any of my patches landed now. Not sure why this is happening.
Reporter | ||
Comment 12•3 years ago
|
||
(In reply to Zaggy1024 from comment #11)
In the bug description you mentioned "It is also worth to notice that when you switch to video, it starts to play with delay." Has the delay reduced in the current Nightly build? The patch that improved WMF decoder init time should have landed that revision, so I'm curious whether it's improved that aspect of the bug.
Looks like it improved slightly but it really hard to certify this.
(In reply to Zaggy1024 from comment #11)
It may be very tricky to nail down this bug, though, since most of the behavior related to H264 should essentially the same as before any of my patches landed now. Not sure why this is happening.
I can help with testing test builds if it will be needed, as I am seeing that in regression range there are 3 patches which might cause the issue.
Comment 13•3 years ago
|
||
Come to think of it, this issue may be related to Bug 1763680. You are normally running Firefox off of your GTX 750 Ti, correct? It's possible that you aren't getting hardware decode anymore on your card.
If you grab a Media profile, it'll give an idea of whether my suspicion is correct, but once a patch lands later today we may be able to request more information from you or the other reporter there to determine the cause.
Comment 14•3 years ago
|
||
https://treeherder.mozilla.org/jobs?repo=autoland&revision=68722c984f71c9adb23fba5405a47c5f05a9266b&selectedTaskRun=IBoOCktrTsahIcYvVUtKZA.0
Builds have finished with the extra logging, you could try "Windows 2012 x64 opt B" -> "Artifacts and Debugging Tools" -> target.zip to get those Media logs. If it's the same issue as Bug 1763680 then we can hopefully figure out what's going on from that.
Comment 15•3 years ago
|
||
The logging patch has also landed in Nightly now, so you can just download the latest Nightly instead to grab the profile.
Reporter | ||
Comment 16•3 years ago
|
||
Yes, I'm using only NVIDIA GTX 750 Ti as GPU.
Done, please have a look at the profile - https://share.firefox.dev/38CYmkP
Comment 17•3 years ago
|
||
Yep, looks like you're not getting hardware decoding due to the same issue as Bug 1763680. I'll mark this as duplicate, hopefully we can get the patch rolled out soon for this.
Updated•3 years ago
|
Updated•3 years ago
|
Description
•