[Live-devel] Live555 RTSP Client never sees RTCP BYE message from Live555 Server

[Matt Schuckmann]

You have inadvertently pointed out something we missed and that is how 
you *intended* the server should force a shutdown of an active session 
or a live feed. We had been simply destroying the RTSPClientSession 
which is what happens when a TEARDOWN message has been received. And 
this is how I discovered the aforementioned problem.

By back tracing to see how afterPlayStreamState() could have been called 
I see that we should have been calling handleClosure() on each of the 
FramedSources, that would cause the fNoFramesLeft flag in 
MultiFramedRTPSink to be set to true and cause the subsession to end, as 
you intended. I will modify our code to do this when the server is 
forcing shutdown of a session.

I still believe that a BYE message should be sent from the server when 
the client initiates the session ending via a TEARDOWN (or other such) 
message.

Matt S.


Ross Finlayson wrote:
>> I'm pretty sure that this would happen with unmodified code, just see 
>> if your bye handler is called in OpenRTSP after sending or a teardown 
>> message while the source is still playing.
>
> It is, *provided that* the original source file does not have a known 
> duration.  When streaming from a live source, or from a file for which 
> the server does not know the duration, the server sends a RTCP "BYE" 
> packet when the input source ends.
>
> However, when streaming from a file for which the server knows the 
> file's duration (currently, MP3, .MPG, .TS and .AVI files), the server 
> *does not* send a RTCP "BYE" packet when it reaches the end of the 
> file.  The reason for this is to keep the stream alive, and therefore 
> allow the client - if it desires - to continue playing the file, by 
> seeking back within it and replaying from an earlier time. Because - 
> in this case - the client was told the stream's duration beforehand, 
> it doesn't need to see a RTCP "BYE" in order to know when it ends.
>
> Note the code for the function "afterPlayingStreamState()" in 
> "liveMedia/OnDemandServerMediaSubsession.cpp".
>
> And yes, I've tested this on the installed code.
>
> If you suspect a bug in the code, then please test this using the 
> original, released code; not your modifications to the code.