Re: [Linphone-developers] Possible issue with the RTP recv timestamps treatment in the 0.12.2 code tree...

[Dave Harley]

Thanks Simon, 
I'll move to 1.0.x and see if I can reproduce the issue and work on it
from there.

Best regards, 
Dave.

On Tue, 2005-04-19 at 10:53 +0200, Simon Morlat wrote:
> Hello,
> 
> The problem of received packet being out of time is usually a problem with 
> the 
> sender, not the receiver. The sender may sent too fast, or too slowly, 
> resulting in such behaviours. Usually those sender problems are related to 
> soundcard problems. Using alsa driver everywhere solves it.
> There have been progresses in 1.0.x to allow linphone to compensate slow 
> clock 
> slides between sender and receiver. This may help.
> Simon
> 
> Le Jeudi 14 Avril 2005 20:23, Dave Harley a écrit :
> > Hi there,
> > Sorry to post out of the blue but I'm having an issue with received
> > audio data in the rtp stream.  Any input greatly welcome...
> >
> > The symptom is that I lose audio in the speaker; it goes quiet.  The
> > "Mic" side is fine and the called party (remote) telephone receives its
> > audio perfectly.  The called party also seems to transmit fine and the
> > packets are seen on the wire and with good headers.  The setup is
> > linphone 0.12.2 on a 1 Mhz Via machine over a cross-over cable (I use a
> > hub when taking line traces) to a standalone sip phone.  100 base T
> > ethernet full duplex (at least on the crossover).
> >
> > Investigation so far:  Traced the issue up the stack from the ALSA API
> > through the mlaw decoder and down to the RTP receiver.  A merry dance.
> > It appears that my rtp recv queue is going empty and that when in this
> > state it remains empty until a RESYNC occurs.  At that point audio
> > returns for a few seconds, then the queue returns to the same state
> > <empty> as before and audio is gone again.  Packets are available from
> > the socket all through this sequence.
> > If I do not call:
> >   rtp_session_unset_flag (session, RTP_SESSION_RECV_SYNC);
> >   in rtp_session(...)
> > then all works fine.  I know that this is not a fix but its interesting
> > that ignoring the time_stamps seems to get around the issue.  Clear
> > bidirectional audio.
> >
> > So underlying all this seems to be the user_ts (time-stamp) getting out
> > of sync with the rpt's time-stamp.  In this state the rtp_parse function
> > throws the packet away as being "too old" and so never they never get to
> > the rtp_recv queue.  I had a look at the 1.0.1 code and there seems to
> > have been a lot of work in this area.  Is this a known issue in the
> > 0.12.2 build?
> >
> > Anyway thanks in advance for any input.  Hope my description is in some
> > way clear.
> >
> > Best regards,
> > Dave
> >
> >
> >
> >
> > _______________________________________________
> > Linphone-developers mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/linphone-developers
>