[jitsi-dev] Where in the bridge does it decide to send a translated RTP packet


I have pinned down the issue of data not arriving at Chrome to conclude data is being sent to the Bridge by Chrome, this is being properly allocated to the correct RtpChannel/Video/Audio whatever. However, then the bridge appears to simply discarding it rather than forwarding it on.

My earlier point about the remote SSRCs not being allocated in the bridge is probably linked to this. I will have a go at cross linking those when a new SSRC is added in RtpChannel so that the prior description of the channel is updated.

Additionally there appears to be an issue with having more than one connection from the same host in that the first DTLs connection works well, but subsequent ones sometimes work first time and sometimes need a few goes. The logging message is:

INFO: Pair failed: [2001:0:5ef5:79fd:3425:391c:2b60:fab3]:10000/udp/host -> [2001:0:5ef5:79fd:3425:391c:2b60:fab3]:10000/udp/host (stream.RTP)
Jan 22, 2017 5:50:55 PM org.ice4j.ice.harvest.SinglePortUdpHarvester maybeAcceptNewSession
INFO: Failed to handle new socket: java.io.IOException: Agent state is Terminated. Cannot add socket.


Question: Where should I look for the decisionmaking as to what to do with the datagrams that should be translated, but are not being. I think it is something to do with a LIstener relating to an instance of A Mediastream class (or derivative)?