Lennart,
Your code loop appears to have resolved the issues between asterisk and pulse in the write direction, now I have to
resolve the read direction. Can you provide some guidance on a preferred processing loop for reading from pulse?
I took a stab, how does this look?
for (;;) {
state = snd_pcm_state(alsa.icard);
if ((state != SND_PCM_STATE_PREPARED) && (state != SND_PCM_STATE_RUNNING)) {
snd_pcm_prepare(alsa.icard);
}
if (state == SND_PCM_STATE_XRUN)
snd_pcm_prepare(alsa.icard);
buf = __buf + AST_FRIENDLY_OFFSET / 2;
r = snd_pcm_readi(alsa.icard, buf + readpos, left);
if (r >= 0)
/* Success */
off -= r;
break;
/* Failure, try to recover */
if ((r = snd_pcm_recover( alsa.icard, r, 1)) < 0) {
/* Failed to recover, we have a real issue */
ast_log(LOG_NOTICE, "snd_pcm_recover failed: %s", snd_strerror(r));
break;
}
/* Recovered, let's try another read */
}
"Lennart Poettering" <lennart at poettering.net> wrote in message news:20080215221012.GA18000 at tango.0pointer.de...
> On Thu, 14.02.08 20:10, Jim Duda (jim at duda.tzo.com) wrote:
>
>>
>> Here is the Asterisk alsa write code loop. I added the trap for -EIO.
>> I'm getting an negative return from snd_pcm_recover.
>>
>> [Feb 14 20:08:25] ERROR[3085]: chan_alsa.c:404 alsa_write:
>> snd_pcm_recover failed Input/output error
>>
>> Is there anything obvious which needs to be done to make this more pulse
>> friendly?
>
> Not sure. The general logic of snd_pcm_recover should be like this:
>
> for (;;) {
> if ((err = snd_pcm_writei(....)) >= 0)
> /* Success */
> break;
>
> /* Failure, try to recover */
> if ((err = snd_pcm_recover(..., err)) < 0) {
>
> /* Failed to recover, we have a real issue */
> fprintf(stderr, "Things went very wrong: %s", snd_strerror(err));
> break;
> }
>
> /* Recovered, let's try another write */
> }
>
> This probably won't fix your issue, but it will shorten your code
> quite a bit I would say and make it more robust.
>
> snd_pcm_prepare() should be called only once after
> init. snd_pcm_recover() will implicitly call _prepare() in all other
> cases.
>
> Lennart
>
> --
> Lennart Poettering Red Hat, Inc.
> lennart [at] poettering [dot] net ICQ# 11060553
> http://0pointer.net/lennart/ GnuPG 0x1A015CC4
