[alsa-devel] snd_dummy on Centos - Redhat

Takashi Iwai tiwai at suse.de
Fri Sep 4 12:26:08 CEST 2009


At Fri, 04 Sep 2009 12:18:10 +0200,
I wrote:
> 
> At Fri, 4 Sep 2009 11:33:32 +0200,
> Giovanni Maruzzelli wrote:
> > 
> > On Fri, Sep 4, 2009 at 11:20 AM, Takashi Iwai<tiwai at suse.de> wrote:
> > >
> > > If it's XRUN, the stream has to be stopped once, so you must have
> > > STOP log there.
> > >
> > >> So, this time for not having false positives I send you the result of
> > >> a continuous play of music, without pauses in it (maybe a pause at the
> > >> beginning):
> > >
> > > Hrm, is this the output from one stream?
> > >
> > >> Sep  4 11:04:04 localhost kernel: *** STOP
> > >> Sep  4 11:04:25 localhost kernel: *** START: buffer=4096, period=2048,
> > >> rate=16000
> > >> Sep  4 11:04:25 localhost kernel: xxx delta=2, pos=32000(32), rest=2016000(2016)
> > >> Sep  4 11:04:25 localhost kernel: xxx delta=3, pos=80000(80), rest=1968000(1968)
> > >> Sep  4 11:04:25 localhost kernel: *** START: buffer=4096, period=2048,
> > >> rate=16000
> > >
> > > Here you have the second start without stopping.
> > >
> > 
> > For what I have seen hacking out my modifications to the dummy
> > drivers, Skype open and close a lot the devices... Probably it do that
> > for "ringing", making other announcements, etc.
> > 
> > They use the RTaudio library, probably with their own modifications....
> > 
> > Also, maybe the second START is the start of capture...?
> 
> Ah, of course.  I always forget you are debugging Skype :)
> Maybe better to add a substream id (e.g. substream->stream,
> substream->number) to avoid confusion.
> 
> If they are different substreams, my analysis doesn't make sense at
> all.  If so, I don't see any particular problem in the log...

BTW, I added another patch to snd-dummy driver in sound-unstable tree.
Now it doesn't allocate the real buffers but fakes.  It would save
lots of memory when you run lots of substreams.


Takashi


More information about the Alsa-devel mailing list