Anders Broman wrote:
I have checked in the plugin files for now there are some problems
With applying the change to the makefiles can you please redo the
Patch against the current SVN. I would also suggest to leave
"PORTAUDIO" commented out in these first checkins.
There are also some warnings from voip_calls.c
At least part of the problem is that the dissector calling mechanism is
being hijacked to call what's really more like a tap.
I have a version with the listen_rtp stuff in the gtk directory as
builtin code rather than as a plugin, so it doesn't use the dissector
calling mechanism in that fashion, and with configure-script and nmake
file conditionalization so it should build with or without PortAudio.
It builds on OS X.
However, it doesn't actually play anything on the sample capture.
It appears that the problem is that, at least for the call in the
capture, which was set up with H.323, only the setup packets are treated
as part of the call, not the RTP packets - the graph for the call shows
only the H.225 and H.245 packets.
This means that mark_rtp_stream_to_play() will not find
rsi->first_frame_number anywhere in the list of packets in the graph,
and thus won't mark the stream as a stream to be played.
Is it intentional that only the setup calls, not the voice traffic, are
in the graph?