Launchpad has imported 21 comments from the remote bug at
https://bugzilla.mozilla.org/show_bug.cgi?id=1645671.

If you reply to an imported comment from within Launchpad, your comment
will be sent to the remote bug automatically. Read more about
Launchpad's inter-bugtracker facilities at
https://help.launchpad.net/InterBugTracking.

------------------------------------------------------------------------
On 2020-06-14T11:26:27+00:00 W-jan-k wrote:

Created attachment 9156561
Bildschirmfoto vom 2020-06-14 12-57-09.png

This is a minor cosmetic issue. It affects software and VAAPI.
It occurs only once. To reproduce again, reload the tab or change video 
resolution on YouTube.
(With media.ffvpx.enabled:false this can be observed with VP9 videos.)

software: MOZ_LOG="PlatformDecoderModule:5" MOZ_ENABLE_WAYLAND=1
mozregression --launch 0c7df6f9b0c1 --pref gfx.webrender.all:true widget
.wayland-dmabuf-video-textures.enabled:true media.ffvpx.enabled:false -a
https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605 -P
stdout

hardware: MOZ_LOG="PlatformDecoderModule:5" LIBVA_DRIVER_NAME=i965
MOZ_ENABLE_WAYLAND=1 mozregression --launch 0c7df6f9b0c1 --pref
gfx.webrender.all:true widget.wayland-dmabuf-vaapi.enabled:true widget
.wayland-dmabuf-video-textures.enabled:true media.ffvpx.enabled:false -a
https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605 -P
stdout

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/0

------------------------------------------------------------------------
On 2020-06-15T06:58:17+00:00 Marko Popović wrote:

I can confirm this bug, started happening a few days ago.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/1

------------------------------------------------------------------------
On 2020-06-16T02:13:56+00:00 甘露 (Lu Gan) wrote:

I meet the similar issue with huya.com which is a live streaming
website.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/2

------------------------------------------------------------------------
On 2020-06-16T13:23:11+00:00 Stransky wrote:

So this is a regression from Bug 1629788, correct?

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/3

------------------------------------------------------------------------
On 2020-06-16T13:30:02+00:00 W-jan-k wrote:

Correct.
https://hg.mozilla.org/integration/autoland/shortlog/0c7df6f9b0c1
first affected build: MOZ_LOG="PlatformDecoderModule:5" MOZ_ENABLE_WAYLAND=1 
mozregression --repo autoland --launch 0c7df6f9b0c1 --pref 
gfx.webrender.all:true widget.wayland-dmabuf-video-textures.enabled:true 
media.ffvpx.enabled:false -a 
https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605 -P stdout
last unaffected build: MOZ_LOG="PlatformDecoderModule:5" MOZ_ENABLE_WAYLAND=1 
mozregression --repo autoland --launch 6f6757bfd3298d8314d52730c3c016dc4c629f05 
--pref gfx.webrender.all:true widget.wayland-dmabuf-video-textures.enabled:true 
media.ffvpx.enabled:false -a 
https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605 -P stdout

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/4

------------------------------------------------------------------------
On 2020-06-16T13:32:55+00:00 W-jan-k wrote:

Same for VAAPI:
first affected build: MOZ_LOG="PlatformDecoderModule:5" LIBVA_DRIVER_NAME=i965 
MOZ_ENABLE_WAYLAND=1 mozregression --launch 0c7df6f9b0c1 --pref 
gfx.webrender.all:true widget.wayland-dmabuf-vaapi.enabled:true 
widget.wayland-dmabuf-video-textures.enabled:true media.ffvpx.enabled:false -a 
https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605 -P stdout
last unaffected build: MOZ_LOG="PlatformDecoderModule:5" LIBVA_DRIVER_NAME=i965 
MOZ_ENABLE_WAYLAND=1 mozregression --repo autoland --launch 
6f6757bfd3298d8314d52730c3c016dc4c629f05 --pref gfx.webrender.all:true 
widget.wayland-dmabuf-vaapi.enabled:true 
widget.wayland-dmabuf-video-textures.enabled:true media.ffvpx.enabled:false -a 
https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605 -P stdout

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/5

------------------------------------------------------------------------
On 2020-06-18T10:37:44+00:00 Marko Popović wrote:

This started happening again after today's update...

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/6

------------------------------------------------------------------------
On 2020-06-18T11:05:48+00:00 W-jan-k wrote:

This was never gone and the feature is not finished yet (bug 1629788 comment 
27). Some other patches still await review.
VAAPI was just temporarily disabled: bug 1645706 disabled the 
dmabuf-video-textures pref that VAAPI now depends upon. You had to enable two 
prefs for VAAPI. bug 1646051 changed it to let the VAAPI pref also enable 
dmabuf-video-textures. You see this bug again because you are using VAAPI again.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/7

------------------------------------------------------------------------
On 2020-06-18T11:57:53+00:00 Marko Popović wrote:

(In reply to Jan Andre Ikenmeyer [:darkspirit] from comment #7)
> This was never gone and the feature is not finished yet (bug 1629788 comment 
> 27). Some other patches still await review.
> VAAPI was just temporarily disabled: bug 1645706 disabled the 
> dmabuf-video-textures pref that VAAPI now depends upon. You had to enable two 
> prefs for VAAPI. bug 1646051 changed it to let the VAAPI pref also enable 
> dmabuf-video-textures. You see this bug again because you are using VAAPI 
> again.

Thanks for clarification! I thought it was gone for a few days.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/8

------------------------------------------------------------------------
On 2020-06-18T14:52:13+00:00 George Sofianos wrote:

(In reply to Marko from comment #8)
> Thanks for clarification! I thought it was gone for a few days.

I never had this on my PC (AMD 5700XT / mesa 20.1.1 / Linux 5.7.2) and I
was using VAAPI decoding for about a month. Just started happening
today. Even with yesterdays nightly where I had to enable the dmabuf-
video-textures pref, I was seeing no Green frames. I'm just commenting
it in case it helps to identify what it is, but since it's work in
progress I guess it will be resolved at some point.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/9

------------------------------------------------------------------------
On 2020-07-05T15:58:56+00:00 W-jan-k wrote:

*** Bug 1650648 has been marked as a duplicate of this bug. ***

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/10

------------------------------------------------------------------------
On 2020-07-15T12:49:08+00:00 O-mj-q wrote:

For reference, this is also an issue on X11/EGL, even without
`media.ffmpeg.dmabuf-textures.enabled`.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/11

------------------------------------------------------------------------
On 2020-07-23T14:26:31+00:00 Marko Popović wrote:

Why is this bug of such low priority, it is a regression that basically
isn't present on 78 stable and will regress both 79 and even 80 it seems
for all the vaapi users... is there any way that fix will be backported
once it's fixed?

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/12

------------------------------------------------------------------------
On 2020-07-28T12:06:21+00:00 Pmargeti34 wrote:

I agree with comment 12, now that firefox 79 is officially out this bug
is very visible to anyone using hardware accelerated video playback on
linux.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/13

------------------------------------------------------------------------
On 2020-07-28T12:38:27+00:00 Robert Mader wrote:

(In reply to Marko from comment #12)
> Why is this bug of such low priority, it is a regression that basically isn't 
> present on 78 stable and will regress both 79 and even 80 it seems for all 
> the vaapi users... is there any way that fix will be backported once it's 
> fixed?

First of all this feature is not yet enabled by default and requires
users to jump through several hoops to enable it (enable EGL either for
X11 or by using the Wayland backend, enable WR, enable several options).
Secondly it's not a crash but just a glitch, and not a major one
neither. Finally the feature was not developed by the FF core team but
external contributors. This together results in a low priority. And no,
fixes for disabled-by-default features will usually not get backported.

But don't worry, I'm sure Martin will work on it as time allows.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/14

------------------------------------------------------------------------
On 2020-07-28T12:47:30+00:00 Marko Popović wrote:

(In reply to Robert Mader [:rmader] from comment #14)
> (In reply to Marko from comment #12)
> > Why is this bug of such low priority, it is a regression that basically 
> > isn't present on 78 stable and will regress both 79 and even 80 it seems 
> > for all the vaapi users... is there any way that fix will be backported 
> > once it's fixed?
> 
> First of all this feature is not yet enabled by default and requires users to 
> jump through several hoops to enable it (enable EGL either for X11 or by 
> using the Wayland backend, enable WR, enable several options). Secondly it's 
> not a crash but just a glitch, and not a major one neither. Finally the 
> feature was not developed by the FF core team but external contributors. This 
> together results in a low priority. And no, fixes for disabled-by-default 
> features will usually not get backported.
> 
> But don't worry, I'm sure Martin will work on it as time allows.

I know, but it's a regression... shouldn't regressions have higher priority 
than implementing new features like X11 backend etc... the part that is bad 
about this is that in 78 it works perfectly without any issues, and in 79+ is 
annoying to use again just like with frame skipping bug before, and yet there 
are new features being worked on without finishing old ones first. 
Obviously it's up to Martin to decide since he is a contributor and FF devs 
don't care enough to provide tge feature, but as a linux laptop user, I would 
really like to see this having a higher priprity :) didn't mean to be rude or 
anything.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/15

------------------------------------------------------------------------
On 2020-07-28T12:57:33+00:00 Stransky wrote:

*** Bug 1645673 has been marked as a duplicate of this bug. ***

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/16

------------------------------------------------------------------------
On 2020-07-28T13:05:00+00:00 Stransky wrote:

(In reply to Marko from comment #15)
> I know, but it's a regression... shouldn't regressions have higher priority 
> than implementing new features like X11 backend etc... the part that is bad 
> about this is that in 78 it works perfectly without any issues, and in 79+ is 
> annoying to use again just like with frame skipping bug before, and yet there 
> are new features being worked on without finishing old ones first. 
> Obviously it's up to Martin to decide since he is a contributor and FF devs 
> don't care enough to provide tge feature, but as a linux laptop user, I would 
> really like to see this having a higher priprity :) didn't mean to be rude or 
> anything.

The patch is on the way. I prefer to unify X11/Wayland VA-API/dmabuf
backends and then do all the fixes for both together, that lowers the
maintenance cost. This is really still an experimental feature so please
accept potential breakage during the development (and I also have other
duties at Red Hat).

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/17

------------------------------------------------------------------------
On 2020-07-28T13:10:00+00:00 Marko Popović wrote:

(In reply to Martin Stránský [:stransky] from comment #17)
> (In reply to Marko from comment #15)
> > I know, but it's a regression... shouldn't regressions have higher priority 
> > than implementing new features like X11 backend etc... the part that is bad 
> > about this is that in 78 it works perfectly without any issues, and in 79+ 
> > is annoying to use again just like with frame skipping bug before, and yet 
> > there are new features being worked on without finishing old ones first. 
> > Obviously it's up to Martin to decide since he is a contributor and FF devs 
> > don't care enough to provide tge feature, but as a linux laptop user, I 
> > would really like to see this having a higher priprity :) didn't mean to be 
> > rude or anything.
> 
> The patch is on the way. I prefer to unify X11/Wayland VA-API/dmabuf backends 
> and then do all the fixes for both together, that lowers the maintenance 
> cost. This is really still an experimental feature so please accept potential 
> breakage during the development (and I also have other duties at Red Hat).

Thnaks for the clarification about X11/wayland common backend.

Don't worry Martin, we owe you eternal gratitude for implementing this
feature in Firefox after so many years the issue was neglected by other
developers!

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/18

------------------------------------------------------------------------
On 2020-07-28T13:23:31+00:00 Stransky wrote:

Created attachment 9166559
Bug 1645671 [Linux/VA-API] Create DMABufSurfaceWrapper directly at nsTAttay and 
disable DMABufSurfaceWrapper class copy and assignment constructors, r?jya


When DMABufSurfaceWrapper is added to nsTArray, a temporary local 
DMABufSurfaceWrapper object is created. When the temporary
object is deleted after the adding, associated dmabuf data is released which 
leads to rendering artifact during VA-API video playback.

As a fix in this patch we use UniquePtr<DMABufSurfaceWrapper> at nsTArray thus 
we don't create the temporary DMABufSurfaceWrapper object.
Also change GetUnusedDMABufSurfaceWrapper() to 
GetUnusedDMABufSurfaceWrapperIndex() and reference the surface wrapper by 
indexe instead of pointer.

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/19

------------------------------------------------------------------------
On 2020-07-28T13:49:38+00:00 Stransky wrote:

*** Bug 1645672 has been marked as a duplicate of this bug. ***

Reply at:
https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/1889784/comments/20


** Changed in: firefox
       Status: Unknown => In Progress

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1889784

Title:
  Hardware-accelerated video decoding (VA-API) broken in Firefox 79
  (Wayland)

To manage notifications about this bug go to:
https://bugs.launchpad.net/firefox/+bug/1889784/+subscriptions

-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to