[LAU] jack and RT-patched kernels

[Florian Schmidt]

On Thursday 31 January 2008, Julien Claassen wrote:
> Hi folks!
>   OK beat me round the head with a rubber-bat. I gave /dev/shm as jack's
> tmpdir. Not good. Now LS runs fine. BUT!!!
>   But still when I do the following - and this is repeatable:
> start fluidsynth like this:
> fluidsynth -a jack -j -r 48000 nd_bass_guitar.sf2
>   Start ecasound and use it:
> ecasound -B:rtlowlatency -b:64 -r 30 -c -a:1 -i jack_auto,fluidsynth \
>   -o test.wav -a:2 -i null -pn:met,120 -a:3 -i monitor.wav -a:2,3 -o
> jack_alsa
>
>   And then after working finish ecasound. Fluidsynth looses its connection
> because JACK simply dies, saying:
> jackd watchdog: timeout - killing jackd
> Aborted
>   This is not nice, though sufferable. But as sufferable still contains "to
> suffer", it would be nice to abolish it.
>   Can anyone help here?
>   Kindest regards
>         Julien

I'm not sure this is a jack problem. The dependence on the  size of the loaded 
data files rather hints at problems in the clients that only show in realtime 
mode.. Maybe try 

a] allowing apps to mlock more memory
b] using a longer timeout for jack [see manpage]

/dev/shm is actually a rather good choice for the jack temp directory.. [when 
there's a shmfs mounted there]

Flo


-- 
Palimm Palimm!
http://tapas.affenbande.org