762658 - Attached Ogg cannot be played on some Linux configs with cubeb

Status: RESOLVED WORKSFORME

(Core :: Audio/Video, defect)

Description

[Matěj Cepl]

I have regenerated couple of MP3 files with ffmpeg -i file.mp3 -acodec libvorbis file.ogg and the resulting files play perfectly well in ffplay, or in any Gstreamer based player (tried totem and rhtyhmbox), but in Firefox (today's nightly, Mozilla/5.0 (X11; Linux x86_64; rv:16.0) Gecko/16.0 Firefox/16.0a1) I get only noise with it seems to me wrong length of the file.

One possibility would be non-standard sample rate which other applications can handle well.

Comment 1

[Matěj Cepl]

Attachment: sample OGG file

Comment 2

[Matthew Gregan [:kinetik]]

It plays fine on Windows 7 in a current nightly.  Can you try creating the pref "media.use_cubeb", setting it to false, and retesting?

Comment 3

[Ralph Giles (:rillian)]

Plays fine for me on Fedora 17 with current nightly. File passes oggz-validate as well.

Comment 4

[Paul Adenot (:padenot)]

fwiw, works fine on my Linux too (PulseAudio).

Comment 5

[Matěj Cepl]

(In reply to Matthew Gregan [:kinetik] from comment #2)
> It plays fine on Windows 7 in a current nightly.  Can you try creating the
> pref "media.use_cubeb", setting it to false, and retesting?

yes, media.use_cubeb false makes the sound work correctly.

Comment 6

[Matthew Gregan [:kinetik]]

(In reply to Matej Cepl from comment #5)
> yes, media.use_cubeb false makes the sound work correctly.

Okay, so the question is: what's different between you Fedora 17 system and Ralph's.  Probably audio hardware.  Have either of you change the configuration of the audio subsystem compared to a base install (e.g. disabling PulseAudio, or something else)?

Matej, would you mind attaching the output from alsa-info.sh? (Found here: http://git.alsa-project.org/?p=alsa-driver.git;a=blob_plain;f=utils/alsa-info.sh)

Comment 7

[Ralph Giles (:rillian)]

My machine isn't a fresh Fedora 17 install, it's been upgraded from 15 and 16. I am using pulseaudio on top of alsa with the motherboard's built-in audio. I do have jackd installed but haven't used it lately.

alsa-info output for my machine:

  http://www.alsa-project.org/db/?f=34e7271ee844ccc5f6784e9c4a665f2e7b61d1e1

Comment 8

[Matěj Cepl]

Attachment: output of alsa-info.sh

(In reply to Ralph Giles (:rillian) from comment #7)
> My machine isn't a fresh Fedora 17 install, it's been upgraded from 15 and
> 16. I am using pulseaudio on top of alsa with the motherboard's built-in
> audio. I do have jackd installed but haven't used it lately.
> 
> alsa-info output for my machine:
> 
>   http://www.alsa-project.org/db/?f=34e7271ee844ccc5f6784e9c4a665f2e7b61d1e1

Here we are.

Comment 9

[Ralph Giles (:rillian)]

Attachment: output of alsa-info on my working machine

Attaching my for completeness. Same alsa version.

Comment 10

[Matthew Gregan [:kinetik]]

Thanks for the alsa-info.sh output.  You're both using the same driver (Intel HDA), but have different audio codecs that support different sample rates.  Looking at mine, it's similar to Matej's but one generation older (I've got a Thinkpad W510).  I'll try upgrading to Fefora 17 to see if this problem shows up then.

Matej, does audio work at other sample rates? E.g. http://ompldr.org/vZGhjYg/440Hz.ogg is 48kHz.

Would you mind running PulseAudio in verbose mode and posting the log while attempting to play the Ogg you attached?  Instructions here: https://wiki.ubuntu.com/PulseAudio/Log

Comment 11

[Matthew Gregan [:kinetik]]

No luck reproducing it on my W510 now that it has been upgraded to Fedora 17.  Hopefully the PulseAudio log will produce some leads.

Comment 12

[Matěj Cepl]

(In reply to Matthew Gregan [:kinetik] from comment #10)
> Matej, does audio work at other sample rates? E.g.
> http://ompldr.org/vZGhjYg/440Hz.ogg is 48kHz.

I don't know if it is good or bad, but with use.cubeb switch back to default (i.e., it evaporated from the about:config list) and today's nightly Mozilla/5.0 (X11; Linux x86_64; rv:16.0) Gecko/16.0 Firefox/16.0a1 I just cannot reproduce it at all. Whatever files I tried I have a crystal clear sound suddenly. Both the attached sample and your testing sound are both OK. Before that I could reliably reproduce the problem with various sample rates (I expressly did try 44100 Hz).

Comment 13

[Matěj Cepl]

(In reply to Matej Cepl from comment #12)
> I don't know if it is good or bad, but with use.cubeb switch back to default
> (i.e., it evaporated from the about:config list) and today's nightly
> Mozilla/5.0 (X11; Linux x86_64; rv:16.0) Gecko/16.0 Firefox/16.0a1 I just
> cannot reproduce it at all.

I have to add though that some serious upgrade of my system happened in meanwhile, so I cannot swear the issue in Firefox has been fixed.

Comment 14

[Matthew Gregan [:kinetik]]

Just to clarify, a software or hardware upgrade?

I'm glad it's working correctly for you now.  Since I don't have any leads on this and you can no longer reproduce it, I'll go ahead and mark this as WFM and we can reopen if it shows up again for you or anyone else.

Comment 15

[Matěj Cepl]

(In reply to Matthew Gregan [:kinetik] from comment #14)
> Just to clarify, a software or hardware upgrade?

Software ... as you can guess I have a development system here, so things are changing quite a bit underneath the Firefox.

Comment 16

[Matěj Cepl]

I am cautiously reopening this bug, because it happened to me again. After an hour or so playing a movie via totem (without any problem) I started to play a music from my OwnCloud server and the sound was same horrible again. Yes, pulseaudio was then at really high (50% CPU or so, not 100%), but when I restarted it to capture log, it started to work again flawlessly, so I don't have a proper log (I am attaching what I have but without much hope it would be useful).

The reason why I am reopening this bug here and not on bugzilla.redhat.com against pulseaudio is that as I said totem was working without a problem.

Comment 17

[Matěj Cepl]

Attachment: pulseverbose.txt per Ubuntu instructions

Comment 18

[Matthew Gregan [:kinetik]]

Thanks, I'll try reproducing again while trying to induce high CPU in PA and see if anything turns up.  Is the PA log that you attached from a session where the problem occurred?

Comment 19

[Matěj Cepl]

(In reply to Matthew Gregan [:kinetik] from comment #18)
> Is the PA log that you attached from a session where the problem occurred?

Let me repeat:

# when I restarted it to capture log, it started to work again flawlessly, so I don't have a proper log

Comment 20

[Matthew Gregan [:kinetik]]

This might've been fixed by the patch in bug 761274.  That bug was caused by PA requiring a higher minimum latency than Firefox had requested, resulting in a string of underruns.  PA adjusts the latency upwards as necessary, and for a long-running PA instance on a loaded machine it may grow larger than what Firefox requests.  Restarting PA would fix this problem temporarily, as the minimum latency would be reset by the server restart.

Have you seen this bug since bug 761274 landed?

Comment 21

[Matěj Cepl]

(In reply to Matthew Gregan [:kinetik] from comment #20)
> Have you seen this bug since bug 761274 landed?

Is it in the current Aurora? Nightly was so much crashing on me (both on desktop/Linux as well as on Android) so I had to downgrade.

Comment 22

[Matthew Gregan [:kinetik]]

Yeah, the same fix was landed in Aurora just over a week ago.

Comment 23

[Matěj Cepl]

I am not able to reproduce this even with use.cubeb in default value.