Video calls crash out.

[jaduncan]

Binary package hint: empathy

When using the Jaunty Empathy client on both ends, a voice call can be connected between two hosts. The audio is sent correctly. When video is sent from either end, the call instantly disconnects. The cameras concerned are a Lenovo UVC camera from a x200s (Bus 001 Device 003: ID 17ef:480c Lenovo) and a Sony Motion Eye (Bus 001 Device 003: ID 05ca:1839 Ricoh Co., Ltd) from a PCG-5J5M. Both devices display an image in gstreamer-properties, so the input should work. Audio works, so the transfer of data seems ok (it should be noted that both devices are on the same NAT network). The calls are from gtalk to gtalk account, although the bug is also reproducible when creating an account on jabber.org and as mentioned the audio does go through so the connection seems OK. I am confused. I also note that the cameras are both high resolution, but show up in the preview window fine. I have attached a screen shot of this (and why yes, it is me looking like a man filing a bug report). The screenshot is taken at the time after the voice call is created, and the video preview is fine but before the point at which the video call is initiated and the receiving client disconnects. The sending client is not aware that the call has ended at this point (might this be due to a crash on the receiving end rather than a neat ending of the call?), but continues to count time upwards as if the call is fine but it happens to be receiving no data. I will also include the empathy.log from sender and receiver in the comments.
.
The one thing I did wonder is if the 1280*1024 capture resolution of the camera is crashing the stream somehow.

The video failure thus seems like a large bug, and one that I had assumed would be fixed by the 2.26.0.1 updates. Versions of relevant software are current as of Jaunty, 3/Apr/09: empathy 2.26.0.1-0ubuntu1, telepathy-gabble 0.7.22-1ubuntu3, gstreamer-plugins-base 0.10.22-4, and gstreamer-nice 0.0.5-build1

I will be around for all of today, and so any requests for more info can be fulfilled.

[jaduncan (jaduncan)]

Sender's empathy.log.

[jaduncan (jaduncan)]

Receiver empathy.log.

[jaduncan (jaduncan)]

A bit that seems the source of the crash from receiver's empathy.log - is one client really sending a stream that the other client cannot decode? Surely not, given that this is gstrtptheoradepay.c. It looks like an error with getting the two cryptography pads right. This would also appear to imply that the size of the stream is not the issue:

"(empathy:17332): GLib-GObject-WARNING **: IA__g_object_set_valist: object class `GstRTPDTMFMux' has no property named `clock-rate'
** (empathy:17332): DEBUG: stream 2 (video) _tf_stream_bus_message: Send codec changed: 96: video THEORA clock:90000 channels:1 params:0x2494280
** (empathy:17332): DEBUG: stream 2 (video) cb_fs_new_active_candidate_pair: called
** (empathy:17332): DEBUG: tp_call_stream_state_changed_cb: Stream state changed - stream id: 2, state state: 2
** (empathy:17332): DEBUG: stream 2 (video) set_stream_playing: 1
** (empathy:17332): DEBUG: stream 2 (video) tf_stream_request_resource: Requesting resource for direction 2
** (empathy:17332): DEBUG: stream 2 (video) tf_stream_request_resource: Requesting resource for direction 2 returned 1
** (empathy:17332): DEBUG: stream 2 (video) set_stream_sending: 0
** (empathy:17332): DEBUG: stream 2 (video) cb_fs_stream_src_pad_added: New pad
** Message: Element error: Could not decode stream. -- gstrtptheoradepay.c(618): gst_rtp_theora_depay_process (): /GstPipeline:pipeline0/FsRtpConference:fsrtpconference0/GstBin:recv_2_1432799103_96/GstRtpTheoraDepay:rtptheoradepay0:
Could not switch codebooks

** (empathy:17332): DEBUG: tf_channel_dispose
** (empathy:17332): DEBUG: _tf_session_dispose"

[jaduncan (jaduncan)]

Confirmed as a bug by upstream at http://bugzilla.gnome.org/show_bug.cgi?id=577811. I cannot alter the importance to reflect that this feature is currently not working from one empathy client to another, both on Jaunty.

[jaduncan (jaduncan)]

Confirmed that the PPA packages fix this - can the fixes be pushed in before release?

[Guillaume Desmottes (cassidy)]

jadunca: which version did you test? Lucid will ship 2.30 so that should be fine.

[Omer Akram (om26er)]

Upstream task marked fixed I am also gonna mark it fixed. If you face this in the development release of empathy please open a new bug report

[jaduncan (jaduncan)]

Just tested lucid - confirmed that this is now resolved.