Re: [Ekiga-list] Building OPAL with H.263

[Steve Hill <steve nexusuk org>]

On 09/22/2011 11:54 AM, Eugen Dedu wrote:

> > Thanks. Does that mean that for Ekiga 3.3.2 you can just install ffmpeg
> > > 0.6.2 (rpmfusion packages 0.7-0.3.20110612git, so presumably that is
> > ok) and Ekiga will just use it directly without needing Opal to provide
> > H.263+ support?
>
> ekiga 3.3.2 depends on opal 3.10.2 anyway, which works with those codecs.

I'm having some significant problems with Ekiga 3.3.2 on Opal 3.10.2.

1. Opal wouldn't build against ffmpeg-0.7-0.3.20110612git.  This 
appeared to be a problem with a line in mpeg4.cxx:
	m_avpicture->pict_type = 0;
After some googling, the suggested fix appeared to be to add a cast:
	m_avpicture->pict_type = (AVPictureType) 0;
Once I'd done that it built ok.

2. Ekiga 3.3.2 doesn't seem to want to register against my Callweaver 
server (Ekiga 3.3.0 works fine).  I can't really tell if this is Ekiga 
at fault or Callweaver.  I've compared the network traffic between the 
two Ekigas, and the only thing I can really note is that Ekiga 3.3.0 
sends a REGISTER and waits for that to succeed (including extra 
round-trips for authentication) before doing anything else.  Ekiga 3.3.2 
sends REGISTER, SUBSCRIBE and PUBLISH packets all at the same time. 
REGISTER and SUBSCRIBE both have the same sequence number (CSeq=1) - is 
this allowed?  Callweaver appears to respond to 3.3.2's SUBSCRIBE with 
"489 Bad Event", which it didn't do under 3.3.0.

In any case, there's something completely crazy going on at the 
Callweaver end because it _does_ return a "200 OK" in response to the 
register, but doesn't actually record the registration in its peers 
list.  I'm guessing it is getting confused by the SUBSCRIBE packet.

3. The H.264 codec appears to completely ignore the bandwidth limit set 
in Ekiga's configuration.  The limit is set to 300Kbps and (according to 
the stats in Ekiga's status bar) it was trying to send about 7Mbps 
(750KB/s).  The video I got back from an echo test was completely 
trashed, but that is quite possibly due to it massively exceeding my 
ADSL upstream bandwidth and hence dropping most of the traffic.

4. H.263 and H.263-1998 both resulted in me just getting a static image 
when I dialled an echo test.  When I terminated the call, Ekiga bombed 
with a segfault.

I'll try to make time to do some more debugging next week.

--

 - Steve