pulseaudio causing crashing of applications

[Andrew Farris]

Lennart Poettering wrote:
> On Thu, 31.01.08 01:31, Andrew Farris (lordmorgul at gmail.com) wrote:
> 
>> Warren Togami wrote:
>>> Hi folks,
>>> Is anyone else seeing regular crashing of applications like pidgin, 
>>> mplayer or xine caused by pulseaudio?  Did anybody manage to get useful 
>>> backtraces out of this?
>>> Thanks,
>>> Warren Togami
>>> wtogami at redhat.com
>> I just had an issue with rhythmbox locking up while playing through 
>> pulseaudio.  It locked only when I switched to VT1 where root was logged 
>> in, and it immediately stopped playing and froze.  I've got the backtrace 
>> for all threads, but the top few are below.  The rest is rather long so I 
>> won't post it yet.  I haven't had rhythmbox do this before tonight.
>>
>> Usually it will stop playing through the sink device (it is still trying 
>> to) until returning to the login that is running rhythmbox, but this time 
>> it actually froze and did not keep streaming to the sink.
> 
> This is somewhat expected behaviour. Access to the audio device
> follows the active session on the display. I.e. if you switch VTs than
> ConsoleKit+HAL will change the ACLs of the audio devices and tell PA
> to stop accessing the audio device. This will cause playback to pause
> for all applications accessing PA. However, as soon as you switch back
> the session, the playback should continue. This is not perfect
> however, since we don't inform Gst/Rhyhtmbox that the audio device is
> temporarily stopped (there's no way to do that afaik).
> 
> Lennart

Well I would expect the writes to the device to be denied, but not that 
rhythmbox would freeze rather than just continuing to try and play.  If in fact 
the applications all received a 'pause' message that would be ok I suppose, but 
that is not quite what happened in that situation.  Rhythmbox actually thought 
it was playing the entire time once or twice, the track time continues, and it 
then gains access to play to the audio device again when returning from VT.  The 
crash is not expected; the track time stopped, and rhythmbox does not respond to 
input.

So I guess this is an area for improvement to get the applications to correctly 
handle this situation since it is designed behavior on pulseaudio's end (to deny 
access while not being the active session).

-- 
Andrew Farris <lordmorgul at gmail.com> www.lordmorgul.net
  gpg 0xC99B1DF3 fingerprint CDEC 6FAD BA27 40DF 707E A2E0 F0F6 E622 C99B 1DF3
No one now has, and no one will ever again get, the big picture. - Daniel Geer
----                                                                       ----