[pulseaudio-discuss] [PACKAGERS] Patches for 0.9.15

[Lennart Poettering]

On Wed, 24.06.09 20:02, Ozan Çağlayan (ozan at pardus.org.tr) wrote:

> 
> Colin Guthrie wrote:
> > 'Twas brillig, and Colin Guthrie at 09/06/09 09:08 did gyre and gimble:
> >> So I've pushed the branch "0.9.15-branch" to my own git clone:
> >> http://colin.guthr.ie/git/pulseaudio/log/?h=0.9.15-branch
> >
> > Lennart Poettering (28):
> >       alsa: allow configuration of fallback device strings in profiles
> >       util: if NULL is passed to pa_path_get_filename() just hand it
> > through
> >       alsa: don't hit an assert when invalid module arguments are passed
> >       alsa: fix wording, we are speaking of card profiles, not output
> > profiles
> >       alsa: initialize buffer size before number of periods to improve
> > compat with some backends
> 
> Okay, this last commit (4bffc7849f7898fa85dd6ac8e735db86b57814b3) broke
> a Creative SB card (snd-ca0106) when PA is launched with tsched=0.
> pavucontrol shows a null output sink instead of a real SoundBlaster one.
> Reverting the commit fixes the issue.
> 
> The problem can be seen on PA -vvv output.

This seems to be a driver issue. Please file a bug against the
driver. What happens is this:

We now set up the hwparams of the device like this:

snd_pcm_hw_params_set_periods_integer()
snd_pcm_hw_params_set_buffer_size_near()  -- which fails, which we log but otherwise ignore
snd_pcm_hw_params_set_periods_near() -- which fails too, which we log but otherwise ignore
snd_pcm_hw_params()   -- which now fails, which is fatal

Previously we did this:

snd_pcm_hw_params_set_periods_integer()
snd_pcm_hw_params_set_periods_near()
snd_pcm_hw_params_set_buffer_size_near()
snd_pcm_hw_params()

In which all of this succeeds.

Under no circumstances a failed call should have the effect of causing
the final snd_pcm_hw_params() call fail.

Lennart

-- 
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net
http://0pointer.net/lennart/           GnuPG 0x1A015CC4