GNOME Bugzilla – Bug 796877
rtpgstpay: Add support for force-keyunit events
Last modified: 2018-08-02 08:09:57 UTC
See commit message
Created attachment 373171 [details] [review] rtpgstpay: Add support for force-keyunit events This triggers immediate re-sending of the configuration data in-band.
Comment on attachment 373171 [details] [review] rtpgstpay: Add support for force-keyunit events Looks reasonable, but should we be using g_atomic_int_*() here seeing that the source event might be coming from a different thread than handle_buffer() runs in?
Created attachment 373172 [details] [review] rtpgstpay: Add support for force-keyunit events This triggers immediate re-sending of the configuration data in-band.
Done, shouldn't matter in practice but is nicer indeed.
Attachment 373172 [details] pushed as 9a80cdb - rtpgstpay: Add support for force-keyunit events
Shouldn't you wait for the force-keyunit even to come back from upstream, either from the encoder or the parser?
Why? Having the caps is useful by itself already without any actual keyframe coming later, and if an actual keyframe comes later then that's just handled as usual. And the event would never come back in many cases, e.g. if you just have "filesrc ! mpegaudioparse ! rtpgstpay ! ...", so those cases wouldn't work then.