[pulseaudio-discuss] questions about no sound, 0 us buffer latency,

[Tanu Kaskinen]

On Wed, 2015-12-02 at 13:40 +0100, Peter Gervai wrote:
> Hello,
> 
> I have been fighting with PA and various SIP desktop clients (and the
> view is rather
> disappointing); it's not your topic anyway.
> 
> However I have been fighting with Linphone with close to no success. I
> am pretty sure that
> they initialise their streams with buggy values but still the result
> may be a bit of PA's fault.
> 
> The effect: I set PA as linphone's output. This results perfect
> capture (mic is working fine) but
> no sound output: total silence. Even for the ringtone test.
> 
> Watching pavucontrol I see that the channel (sink) is opened and that
> it doesn't seem to have
> any data to output. Accidentally, however, I pressed the button with
> "Built-in Audio Analog Stereo"
> and selected the other sink, which is combined for the same, and
> immediately the sound started to be
> hearable.
> 
> After a long session of wild-guessing whats' what in PA I guess I have
> found at least something
> analogous with the problem: when the sound is silent paman (pa
> manager) says that the
> latency is 92777 us which equals to buffer: 0 us + sink: 92777 us.
> This "buffer: 0 µs" didn't look quite
> promising, and indeed when I have changed sink in pavucontrol (from
> normal to combine or vice versa)
> the buffer latency pops up to 30000 us or around and the sound starts.
> 
> My guess is - based on the plenthora of PA problems on the net - that
> the channel was opened with
> some nonsense latency requirement which make PA silent, but why does
> it start to work after switching
> sinks? Maybe PA was sanitising internal stuff?
> 
> Anyway. First, it would be nice to be able to tell the linphone guys
> what they should fix. I'm not
> knowledgeable enough in PA for that.
> 
> Second, maybe this "user error" could be prevented by forcing sane
> values on channels, or log an
> INFORMATIVE warning or something along this line? (I guess that this
> is not a "bug" of PA, but maybe
> it is.)

My guess is that Linphone doesn't initially write any data to the
stream, hence the stream buffer is empty and latency is zero. Moving
the stream to a different sink probably causes the server to request
for more data, and this time Linphone honors the request and audio
starts flowing. But that's just wild guessing.

-- 
Tanu