Red Hat Bugzilla – Bug 494180
Audio stops upon seeking
Last modified: 2009-07-27 19:09:13 EDT
using pulseaudio 0.9.14 (f10) with snd_hda_intel (onboard soundcard) and kernel 22.214.171.124-170.2.56.fc10.x86_64
When trying to seek during a video playing in Miro the sound stops and videoplayback hangs. Using pavucontrol to switch the stream to a different output (usb-soundcard) make the video resume instantly. Restarting pulseaudio also makes the onboard sound work again.
When the "hang" occurs, pulseaudio logs to syslog:
module-alsa-sink.c: ALSA woke us up to write new data to the device, but there was actually nothing to write! Most likely this is an ALSA driver bug. Please report this issue to the PulseAudio developers.
Any known problems with snd_hda_intel? Any fixes for pulseaudio (I found none in the pending 0.9.15-release) or workaround-settings?
The above mentioned case is cleanly reproducible here.
Under some rare circumstances pulseaudio completely dies. Restarting it brings it back to normal. Not sure those cases might be related.
Please run PA in a terminal with -vvvv and paste the debug output which is tehn written to stdout here while you reproduce the issue.
Created attachment 339161 [details]
Output from pulseaudio -vvvv
The problems "E: module-alsa-sink.c: ALSA woke us up to write new data to the device, but there was actually nothing to write! Most likely this is an ALSA driver bug. Please report this issue to the PulseAudio developers." already appeared as soon as the video in Miro started playing. The "Requesting rewind"-things started when I tried to seek, which finally led the video to freeze because soundoutput wasn't possible anymore. As soon as you will the pulseaudio-daemon, Miro reacts again.
Seems the problem is somehow related to adding "tsched=0" to module-hal-detect as described in
Without tsched=0 I don't get the above mentioned, full hangs. But I get audio-skips, especially at the start of a movie.
Hmm, any chance you can get me a full bt over all threads when this happens? i.e. try to reproduce and then attach a gdb to that process with debugging symbols enabled?
Closing due to lack of response to needinfo request.