[alsa-devel] Assert in pcm_params.c

[Alan Horstmann]

On Wednesday 20 May 2015 07:36, Takashi Iwai wrote:
> At Tue, 19 May 2015 20:22:50 +0100,
>
> Alan Horstmann wrote:
> > On Tuesday 19 May 2015 13:01, Takashi Iwai wrote:
> > > At Fri, 1 May 2015 08:36:47 +0100, Alan Horstmann wrote:
> > > > A minimal self-contained demo program ('test-format') has been
> > > > developed and is attached, that demonstrates the issue 100% on the
> > > > reporters machine (HDA-Intel, 6-ch I believe).  The output is:
> > > >
> > > > root at Xeon:/home/patest# ./test-format
> > > > Testing device front
> > > > Num channels 6
> > > > Testing rate: 22050   Result:...Invalid Sample Rate
> > > > Testing rate: 32000   Result:...Invalid Sample Rate
> > > > test-format: pcm_params.c:2249: snd1_pcm_hw_params_slave: Assertion
> > > > `err
> > > >
> > > > >= 0' failed.
> > > >
> > > > Testing rate: 44100   Aborted
> > > >
> > > > Alsa-info is at:
> > > >
> > > > http://www.alsa-project.org/db/?f=19dfeee29f73007e61a00a8fabe3c958f7c
> > > >b8e8 7
> > > >
> > > > This apparently happens with or without Pulseaudio running, with just
> > > > the single 44100 rate, and also with surround devices.  Also, on all
> > > > current Debian and Ubuntu - we have focused on Jessie.
> > > >
> > > > I do not have a machine with similar hardware, so cannot duplicate
> > > > the results.
> > > >
> > > > Any comments, ideas etc would be appreciated.
> > >
> > > I also can't reproduce this.  So this must be pretty specific to the
> > > setup.
> > >
> > > Could you give the exact condition to trigger the problem?  Also, this
> > > happens certainly with the latest alsa-lib?
> >
> > Thanks for taking a look.  Just compiling and running that test program,
> > on the reporter's machine, which has some sort of multi-channel
> > HDA-Intel, is 100% reproducible. The Alsa-lib must be the one provided by
> > Debian Jessie (which was released 25.4.2015); 1.0.28 I think.  Does the
> > Alsa-info give enough details - I can ask any specific questions.
> >
> > However, Raymond seems to have some ideas of a possible cause, in
> > connection with arbitrary period size and softvol; is that plausible?
>
> Possibly, but I wonder why I couldn't see it when I tried with
> different module options.
>
> In anyway, the fix would be simple, the patch like below.  Could you
> check whether it actually fixes your issue?

I've had a positive confirmation that the patch removing the asserts does fix 
the 'crash' in the reporter's system.  So instead it is simply a failure of 
the snd_pcm_hw_params() call, which ultimately can be handled in the users 
code.

I do wonder (as does Raymond) why this fails.  Is 6-chans expected to fail on 
'front'?  In particular the earlier setting of channels/rate/format does not 
fail?  Perhaps I will ask for RULES_DEBUG enabled?

Regards

Alan