Pulseaudio not restarting after crash

Lubomir Kundrak lkundrak at redhat.com
Fri Oct 12 22:54:05 UTC 2007


On Sat, 2007-10-13 at 00:29 +0200, Lennart Poettering wrote:
> On Wed, 10.10.07 06:18, Hikaru Amano (kagesenshi.87 at gmail.com) wrote:
> 
> > [kagesenshi at Hikari ~]$ while true;do pulseaudio ;echo "DIED";done
> > Soft CPU time limit exhausted, terminating.
> > E: module-volume-restore.c: failed to open file '(null)': No such file
> > or directory
> > DIED
> > 
> 
> If the daemon is stopped due to a time limit exhaustion you either
> managed to trigger some kind of endless loop in PA. Or your CPU is
> very, very slow.

Not quite. In my case, I was often experiencing this, but for some
unknown reason I can not reproduce it anymore. Nevertheless, I compiled
pa with -pg -DPRINT_CPU_LOAD to inspect what really happens. While
profiling revealed no useful information (probably it's not pa who's
guilty for the consumption but some of the libs?), PRINT_CPU_LOAD
reveals some interesting information: While I am playing music from
rhythmbox and feed it into pa via ESound emulation, the average CPU
consumption is around 5%. Quite much, as my machine is 2x2200 MHz
(though might be a bit slower due to cpuspeed, maybe 2x800 MHz). When
using native pulseaudio sink, cpu consumption is around 10%.

It is worth noting that CPU load arises suddenly when changing tracks --
I can probably even force PA death by high CPU consumption then.

Until I do some investigate what really consumes the time (BTW how do I
tell oprofile to use debugging symbols from another file as the
executable) -- Does anyone else observe this behavior?

> Could you please run pa with the "-vv" arugment?
> 
> Lennart

-- 
Lubomir Kundrak (Red Hat Security Response Team)




More information about the fedora-devel-list mailing list