[LAU] jack problem getting out of hand

Dragan Noveski perodog at gmx.net
Wed Jul 11 13:49:43 EDT 2007


Tanu Kaskinen wrote:
> On Wed, Jul 11, 2007 at 03:18:42PM +0200, Dragan Noveski wrote:
> [snip]
>   
>> switching on the acpi in general, but switching off the other acpi stuff 
>> like fan, button etc. resolved some issues here.
>>     
> [snip]
>
> Thanks for the tip, I'll have to recompile my kernel again.
>
> But still I don't think this is the cause for Julien's
> problem. I was having very similar symptoms after installing
> the latest jack from svn last week. 
you might be right, i had the same when i tried sth like 0.107.0 from 
svn, so i changed back to 0.103.0!
> At the same time a new
> version of gstreamer0.10-plugins-bad got into Debian
> testing, with the new jackaudiosink. The gstreamer update
> came a bit bad time, because now I don't know if the
> frequent zombification, time-outing and watchdog barking is
> due to some new jack bug or the gstreamer applications
> constantly entering and leaving the processing graph,
> triggering some old misfeature (it seems that the time-outs
> can happen at random times, but I'm not sure, because mostly
> they happen on graph changes and on my system the graph can
> change unexpectedly due to a cron job).
>
> Anyway, my solution was to add a --timeout parameter (with
> value 4000),
it is set here to 500ms, i ll give it a try to complie from svn and 
increase the value.

cheers,
doc
>  and now jackd has been running happily for a
> day without the watchdog coming and killing everyone in
> sight. Pulseaudio and Ingen has been mysteriously zombified
> only once, as opposed to every 15 minutes, and everything
> else has been working fine, although there haven't been much
> stress on the system during the last day.
>
> A related question: what does this error message exactly mean?
>
> subgraph starting at <client> timed out
> (subgraph_wait_fd=<number>, status = 0, state = Running)
>
> Every client is only connected directly to the sound card
> playback ports, so the subgraph shouldn't be particularly
> long. Does the error mean that the said client didn't return
> >from its processing function fast enough or what? After that
> message jack used to stop delivering audio and then after a
> few seconds the watchdog would come and eat everyone. I
> still have got that error twice (the Pulseaudio and Ingen
> zombifications), but the whole jackd hasn't died anymore.
>
>   




More information about the Linux-audio-user mailing list