Re: [fluid-dev] Possible freeze in delete_fluid_alsa_audio_driver()

[David Henningsson]

On 2010-11-03 15:15, Sven Meier wrote:
> Hi Pedro,
>
> thanks for your tips on pulseaudio.
>
> Regarding the freeze, I was able to narrow down the problem.
>
> If the following prerequesites are met:
> - Jack is running,
> - Fluidsynth is explicitely started with -a alsa
> Then the call to snd_pcm_writei() in fluid_alsa.c (line #511) fails to
> return on the 178th invocation - strange!
> The reported freeze on delete_fluid_alsa_audio_driver() is just an
> aftereffect :(.
>
> Seems this isn't a Fluidsynth problem after all.

There was a report recently on the alsa-devel and pulseaudio-discuss 
mailinglists about snd_pcm_writei not returning after a pause (when used 
with the pulse plugin). Perhaps this is the same bug or a related one.

> Can anyone give me an advice where to address this problem?

A simplified way of doing the same as Pedro suggests is to set the 
option audio.alsa.device to plughw:CARDNAME, where CARDNAME is the name 
of your card as listed in /proc/asound/cards. That bypasses pulseaudio, 
grabs exclusive access of the hardware, which gives better performance 
(i e lower latencies) as well.

> Thanks
>
> Sven
>
>
> On 11/03/2010 06:32 AM, Pedro Lopez-Cabanillas wrote:
> > On Tuesday 02 November 2010, Sven Meier wrote:
> > > Hi all,
> > >
> > > I'm experiencing a strange freeze in delete_fluid_alsa_audio_driver():
> > >
> > > 0x9da58ae5 fluid_thread_join + 0x1d
> > > 0x9da4d388 delete_fluid_alsa_audio_driver + 0x47
> > > 0x9da81c74 delete_fluid_audio_driver + 0x59
> > >
> > > The attempt to join the (synthesizer?) thread never returns (line #332):
> > >
> > > dev->cont = 0;
> > > if (dev->thread)
> > > fluid_thread_join (dev->thread); /* never returns */
> > >
> > > To trigger this issue the following prerequisites have to be met:
> > > - Jack is running
> > > - a fluidsynth instance is created with "audio.driver" explicitely set
> > > to "alsa".
> > > - finally the fluidsynth instance is deleted.
> > >
> > > I've just double checked by starting fluidsynth from the command line
> > > and this seems to expose the same issue:
> > >
> > > $ fluidsynth -a alsa
> > > FluidSynth version 1.1.3
> > > ....
> > > fluidsynth: debug: Using 'alsa' audio driver
> > > fluidsynth: debug: Using 'alsa_seq' midi driver
> > > Type 'help' for help topics.
> > > ....
> > > > quit
> > > cheers!
> > > <nothing happens after cheers>
> > >
> > > BTW I'm using the latest fluidsynth from SVN.
> > >
> > > Any ideas?
> > I can confirm the freeze when PulseAudio is also running. Jack typically
> > doesn't use PulseAudio. When I run fluidsynth using an ALSA hw device the
> > freeze doesn't happen. For instance:
> >
> > $ fluidsynth -a alsa -o audio.alsa.device=hw:0,0
> >
> > Note that many cheap audio interfaces don't allow two programs opening
> > the
> > same hw device at the same time.
> >
> > When you don't tell an explicit device, ALSA uses "default" which is
> > routed to
> > the PulseAudio plugin if you have it installed. You can verify this in
> > your
> > "/etc/asound-pulse.conf" or "/etc/asound.conf". To defeat this setting
> > for
> > your own account, you can create a file at your $HOME named
> > ".asoundrc" with a
> > content like this one:
> >
> > pcm.!default {
> > type hw
> > card 0
> > }
> >
> > ctl.!default {
> > type hw
> > card 0
> > }
> >
> > Anyway, the freeze requires more investigation before deciding if it
> > is a bug
> > in ALSA, PulseAudio, or FluidSynth.
> >
> > Regards,
> > Pedro
> >
> > _______________________________________________
> > fluid-dev mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/fluid-dev
>
>
> _______________________________________________
> fluid-dev mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/fluid-dev