bug#44417: ungoogle-chromium crashes on Jitsi & co.

bug#44417: ungoogle-chromium crashes on Jitsi & co.

[Ludovic Courtès]

Hello!

Current ungoogle-chromium (commit
e1f5c2292b88525414b5d0336a00bfa6741d4f14) crashes upon displaying video
as in Jitsi (GPU_DEAD_UPON_ARRIVAL error, IIRC).

Conversely, I found that this one works fine:

--8<---------------cut here---------------start------------->8---
$ /var/guix/profiles/per-user/ludo/current-guix-151-link/bin/guix describe
  guix cd0c426
    repository URL: https://git.savannah.gnu.org/git/guix.git
    branch: master
    commit: cd0c4268d74e58fcfe9c729183d90665f4abbb09
--8<---------------cut here---------------end--------------->8---

Ideas?

Thanks,
Ludo’.

bug#44417: ungoogle-chromium crashes on Jitsi & co.

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 723 bytes --]

Ludovic Courtès <ludovic.courtes@inria.fr> writes:

> Current ungoogle-chromium (commit
> e1f5c2292b88525414b5d0336a00bfa6741d4f14) crashes upon displaying video
> as in Jitsi (GPU_DEAD_UPON_ARRIVAL error, IIRC).

The demo at https://meet.jit.si "works for me".  That is with
"intel-vaapi-driver" installed.  Does that make a difference?

Probably not, because it works here even without $LIBVA_DRIVERS_PATH.

Which GPU are you using?

Recently Chromium started building its own libGL.so and complains about
it not being found at startup (because it looks in $out/lib).  I'm
currently testing a patch that makes it use Mesas libGL mainly to
quiesce the warning.  Not sure if that is related, we'll see...

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 507 bytes --]

bug#44417: ungoogle-chromium crashes on Jitsi & co.

[Ludovic Courtès]

Hi,

Marius Bakke <marius@gnu.org> skribis:

> Ludovic Courtès <ludovic.courtes@inria.fr> writes:
>
>> Current ungoogle-chromium (commit
>> e1f5c2292b88525414b5d0336a00bfa6741d4f14) crashes upon displaying video
>> as in Jitsi (GPU_DEAD_UPON_ARRIVAL error, IIRC).
>
> The demo at https://meet.jit.si "works for me".  That is with
> "intel-vaapi-driver" installed.  Does that make a difference?

Hmm actually Jitsi works for me too.

The one that doesn’t is a private video-conferencing thing at work (not
running Jitsi but very similar).  With the commit above, it says “Error
code: 4” (the GPU error was with an older commit).

The console shows:

--8<---------------cut here---------------start------------->8---
[16695:19:1104/110713.814684:ERROR:audio_rtp_receiver.cc(89)] AudioRtpReceiver::OnSetVolume: No audio channel exists.
[16695:19:1104/110714.263441:ERROR:webrtc_video_engine.cc(3327)] Absent receive stream; ignoring clearing encoded frame sink for ssrc 0
[16695:19:1104/110714.264357:ERROR:webrtc_video_engine.cc(3327)] Absent receive stream; ignoring clearing encoded frame sink for ssrc 0
--8<---------------cut here---------------end--------------->8---

> Probably not, because it works here even without $LIBVA_DRIVERS_PATH.
>
> Which GPU are you using?

It’s i915 and LIBVA_DRIVERS_PATH is unset.

> Recently Chromium started building its own libGL.so and complains about
> it not being found at startup (because it looks in $out/lib).  I'm
> currently testing a patch that makes it use Mesas libGL mainly to
> quiesce the warning.  Not sure if that is related, we'll see...

OK.

Thanks,
Ludo’.

bug#44417: ungoogle-chromium crashes on Jitsi & co.

[Maxim Cournoyer]

Hi Ludo,

Ludovic Courtès <ludovic.courtes@inria.fr> writes:

> Hi,
>
> Marius Bakke <marius@gnu.org> skribis:
>
>> Ludovic Courtès <ludovic.courtes@inria.fr> writes:
>>
>>> Current ungoogle-chromium (commit
>>> e1f5c2292b88525414b5d0336a00bfa6741d4f14) crashes upon displaying video
>>> as in Jitsi (GPU_DEAD_UPON_ARRIVAL error, IIRC).
>>
>> The demo at https://meet.jit.si "works for me".  That is with
>> "intel-vaapi-driver" installed.  Does that make a difference?
>
> Hmm actually Jitsi works for me too.
>
> The one that doesn’t is a private video-conferencing thing at work (not
> running Jitsi but very similar).  With the commit above, it says “Error
> code: 4” (the GPU error was with an older commit).
>
> The console shows:
>
> [16695:19:1104/110713.814684:ERROR:audio_rtp_receiver.cc(89)] AudioRtpReceiver::OnSetVolume: No audio channel exists.
> [16695:19:1104/110714.263441:ERROR:webrtc_video_engine.cc(3327)] Absent receive stream; ignoring clearing encoded frame sink for ssrc 0
> [16695:19:1104/110714.264357:ERROR:webrtc_video_engine.cc(3327)] Absent receive stream; ignoring clearing encoded frame sink for ssrc 0
>
>> Probably not, because it works here even without $LIBVA_DRIVERS_PATH.
>>
>> Which GPU are you using?
>
> It’s i915 and LIBVA_DRIVERS_PATH is unset.
>
>> Recently Chromium started building its own libGL.so and complains about
>> it not being found at startup (because it looks in $out/lib).  I'm
>> currently testing a patch that makes it use Mesas libGL mainly to
>> quiesce the warning.  Not sure if that is related, we'll see...
>
> OK.
>
> Thanks,
> Ludo’.

Just a friendly ping on this issue; at least the title needs to be
updated :-).  What are the '& co'; are there other publicly available
services we can try to reproduce with?

Thanks,

Maxim

bug#44417: ungoogle-chromium crashes on Jitsi & co.

[Ludovic Courtès]

Hi,

Maxim Cournoyer <maxim.cournoyer@gmail.com> skribis:

> Ludovic Courtès <ludovic.courtes@inria.fr> writes:

[...]

>> The one that doesn’t is a private video-conferencing thing at work (not
>> running Jitsi but very similar).  With the commit above, it says “Error
>> code: 4” (the GPU error was with an older commit).
>>
>> The console shows:
>>
>> [16695:19:1104/110713.814684:ERROR:audio_rtp_receiver.cc(89)] AudioRtpReceiver::OnSetVolume: No audio channel exists.
>> [16695:19:1104/110714.263441:ERROR:webrtc_video_engine.cc(3327)] Absent receive stream; ignoring clearing encoded frame sink for ssrc 0
>> [16695:19:1104/110714.264357:ERROR:webrtc_video_engine.cc(3327)] Absent receive stream; ignoring clearing encoded frame sink for ssrc 0

[...]

> Just a friendly ping on this issue; at least the title needs to be
> updated :-).  What are the '& co'; are there other publicly available
> services we can try to reproduce with?

I still experience the same problem with today’s master
(3462678bc346c2f6ea81245d6842264b6dccd945) on that web site (that’s a
Cisco videoconf service).  Unfortunately I don’t know of a public
instance of that service.  :-/

Any ideas of things I could try?

Ludo’.

bug#44417: ungoogle-chromium crashes on some videoconf web sites

[Maxim Cournoyer]

Hi Ludo,

Ludovic Courtès <ludovic.courtes@inria.fr> writes:

> Hi,
>
> Maxim Cournoyer <maxim.cournoyer@gmail.com> skribis:
>
>> Ludovic Courtès <ludovic.courtes@inria.fr> writes:
>
> [...]
>
>>> The one that doesn’t is a private video-conferencing thing at work (not
>>> running Jitsi but very similar).  With the commit above, it says “Error
>>> code: 4” (the GPU error was with an older commit).
>>>
>>> The console shows:
>>>
>>> [16695:19:1104/110713.814684:ERROR:audio_rtp_receiver.cc(89)]
>>> AudioRtpReceiver::OnSetVolume: No audio channel exists.
>>> [16695:19:1104/110714.263441:ERROR:webrtc_video_engine.cc(3327)]
>>> Absent receive stream; ignoring clearing encoded frame sink for
>>> ssrc 0
>>> [16695:19:1104/110714.264357:ERROR:webrtc_video_engine.cc(3327)]
>>> Absent receive stream; ignoring clearing encoded frame sink for
>>> ssrc 0
>
> [...]
>
>> Just a friendly ping on this issue; at least the title needs to be
>> updated :-).  What are the '& co'; are there other publicly available
>> services we can try to reproduce with?
>
> I still experience the same problem with today’s master
> (3462678bc346c2f6ea81245d6842264b6dccd945) on that web site (that’s a
> Cisco videoconf service).  Unfortunately I don’t know of a public
> instance of that service.  :-/
>
> Any ideas of things I could try?

Can you reproduce using the webex test page [0] ?  It works for me.
Although once I had to join a real meeting on it and attempting to share
my screen would crash the browser, IIRC.  I'm using nouveau with an old
nvidia 8800 GTS card.

Maxim

[0]  https://www.webex.com/test-meeting.html#

bug#44417: ungoogle-chromium crashes on some videoconf web sites

[Ludovic Courtès]

Hi Maxim,

Maxim Cournoyer <maxim.cournoyer@gmail.com> skribis:

> Can you reproduce using the webex test page [0] ?  It works for me.
> Although once I had to join a real meeting on it and attempting to share
> my screen would crash the browser, IIRC.  I'm using nouveau with an old
> nvidia 8800 GTS card.
>
> Maxim
>
> [0]  https://www.webex.com/test-meeting.html#

That one does not crash for me with ungoogled-chromium from
c03875b0361f114634caeb54935fe37a9b7b05af.

I’m not sure if there’s any conclusion we can draw?

Thanks,
Ludo’.

bug#44417: ungoogle-chromium crashes on some videoconf web sites

[Maxim Cournoyer]

Hi,

Ludovic Courtès <ludovic.courtes@inria.fr> writes:

> Hi Maxim,
>
> Maxim Cournoyer <maxim.cournoyer@gmail.com> skribis:
>
>> Can you reproduce using the webex test page [0] ?  It works for me.
>> Although once I had to join a real meeting on it and attempting to share
>> my screen would crash the browser, IIRC.  I'm using nouveau with an old
>> nvidia 8800 GTS card.
>>
>> Maxim
>>
>> [0]  https://www.webex.com/test-meeting.html#
>
> That one does not crash for me with ungoogled-chromium from
> c03875b0361f114634caeb54935fe37a9b7b05af.
>
> I’m not sure if there’s any conclusion we can draw?

Are you able to retry the original environment it was not working in?
Without an instance to test, I'm afraid this issue is not actionable :-(.

One thing I've noticed is that ungoogled-chromium sometimes fails to
register with pulseaudio, it seems when pulseaudio was not already
running (launching pavucontrol or having Icecat already running fixes it
for me).  Perhaps it was this?

If you don't have an access to the original reproducer, let's close this
until it comes around again.

Thanks,

Maxim

bug#44417: ungoogle-chromium crashes on some videoconf web sites

[Ludovic Courtès]

Hi Maxim,

Maxim Cournoyer <maxim.cournoyer@gmail.com> skribis:

> Are you able to retry the original environment it was not working in?
> Without an instance to test, I'm afraid this issue is not actionable :-(.

Yeah, it’s not very helpful.  The good thing is that it seems to only
happen on that one site, so it’s not so bad.  Closing!

Thanks for helping out!

Ludo’.