On Mon, Dec 22, 2008 at 11:27:10PM +0100, Pierre Riteau wrote:
> On Mon, Dec 22, 2008 at 09:06:56PM +0000, Jacob Meuser wrote:
> > On Mon, Dec 22, 2008 at 04:22:42PM +0100, Pierre Riteau wrote:
> > > Hi,
> > > I noticed that when running `aucat -l', on some of my machines sndio
> > > apps would play the first sample buffer and then stop (for example,
> > > ogg123, but also sdlmame).
> > > These apps become unkillable by SIGTERM (they are killable by SIGKILL)
> > > until aucat is killed.
> > 
> > I saw this once.  then I realized I had a bunch of different
> > libsndio.so.X.Y, and various of them were being used, depending on
> > how recent things were.  after making sure there was only one
> > libsndio.so and everything was using it, I did not have this problem
> > anymore.
> > 
> > maybe related, maybe not.
> 
> I don't think this is related: I've seen it on fresh installs, and I just
> saw it after cleaning up all old libs in /usr/lib.
> 
> I should note that aucat is killable, but the other apps I tried (only
> ogg123 and sdlmame) cannot be killed.
> 
> > > I found that running `aucat -l -m play` solves this issue, is it normal?
> > > 
> > > It happens on VMware virtual machines but also on a real i386 with an
> > > auich sound chipset (unfortunatly I haven't had the chance to test the
> > > `-m play' workaround on it).
> > > Below is dmesg/audioctl/mixerctl of the VMware VM.
> > > I guess this is because my sound hw doesn't support fullduplex, as
> > > advertised in audioctl (why fullduplex and full_duplex anyway?), but
> > > shouldn't aucat work around this automatically? (I'm just playing sound,
> > > not recording at the same time).
> > 
> > afaik, all auich and eap support full-duplex properly.
> > 
> > $ cd /usr/src/regress/sys/dev/audio_rw
> > $ make DO_AUTEST=1
> > 
> > will run some basic read-only, write-only, and full-duplex tests.
> 
> Yes, I was misled by the fullduplex and full_duplex variables being
> set to 0.

ah, 'audioctl properties' will tell you if the device is supposed to
have full-duplex support.

> It seems it was actually the current state of the device, and as I
> said in my second mail it switched to 1 when using Alexandre's command.
> 
> I tried those regress tests and they run fine without warning or error.
> All I hear is a little cracking between most tests.

ok.  at least basic full-duplex works.  there could be some quirks
wrt blocksize or something though.

-- 
[email protected]
SDF Public Access UNIX System - http://sdf.lonestar.org

Reply via email to