Bug#818083: Makes noise instead of playing music

[Thomas Orgis]

Am Sun, 13 Mar 2016 17:39:38 +0100
schrieb Shérab <Sebastien.Hinderer at ens-lyon.org>: 

> So alsa makes noise, pulse and oss play music and I believe the others
> are not too relevant...

OK, that means that probably the decoding part is fine. Maybe you have
a problem with specific output encodings.

$ mpg123 -vvv

prints a table of encodings (and rates, channel modes) the output
module supports and you also see what choice mpg123 makes.

> Am I correct that the fact that it works with two modules makes this
> test un-necessary?

So pulse and OSS work. I presume you are running Pulseaudio, then?
Interesting that the OSS emulation works and the access via ALSA not.
That reminds me of the times where it was easier to get working audio
through ALSA drivers by using the OSS emulation offered by them. Hm,
are we now dealing with OSS emulated by ALSA which in turn routes to
Pulse?

Or, don't you have pulseaudio normally running and mpg123 starts its
own instance (it does that in the background if no server found)? Then
it's indeed the old game of ALSA OSS emulation working better than ALSA:-/

Verbose mpg123 output might give a hint about problematic settings. You
can enforce an encoding via

$ mpg123 -e $encoding

(see `mpg123 --longhelp |grep encoding` for choices). If things work
with s16, but not with s32, for example, that might be a hint.

If you can turn off pulseaudio / don't have it running, trying hardware
access via ALSA might be fruitful:

$ mpg123 -o alsa -a hw:0,0

The libasound software layers might get something wrong.

> x86_64.
> 
> > 2. Which decoder is used?

So, either the x86-64 SSE or the AVX one should be it. But that is the
same regardless of output method, so …

> So my understanding is that it is the wrong decoder which is chosen,
> right?

… not really. Or, maybe: The output encoding constraints may select a
different decoding path (specific routines for output encodings).

Please re-do a test with

$ mpg123 --cpu generic -o alsa

If that suddenly works, we got probably got an issue with a specific
output format's optimised code path. If not, we can rule out the
decoders and it is really something going wrong later, in transfer. An
off-by-one error in a bytestream makes for nice noise.

But, most importantly: If you noticed that the Pulse output works for
you, you might just use that and forget the weirdness. I gave up
understanding every audio output weirdness. I'm fairly confident in
that mpg123 itself is not to blame. Fairly.


Alrighty then,

Thomas




-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: Digitale Signatur von OpenPGP
URL: <http://lists.alioth.debian.org/pipermail/pkg-multimedia-maintainers/attachments/20160313/6dcc0651/attachment.sig>