fix for peer disconnected.

[jerry]

Hi Arsen,

Have you got a fix for following issue?  

https://groups.google.com/forum/?fromgroups#!searchin/unimrcp/peer$20disconnect/unimrcp/RaCymP0D8SQ/sMPEhl5rujYJ

I think fix for this issue will  solve "unreleased engine channel"problem.   In our system, there are occasional cases where the client will cause  SIP/MRCPv2 connections (from mrcpv2transport's mrcp_server_connection.c, connection agent) to exit ( error is 

An existing connection was forcibly closed by the remote host: 730054).  This will leave the related sofia-session, mrcp-server-session, mrcp-engine-channel, and mrcp-channel and etc unreleased. 

I have traced mrcp code, thinking the following solution. Appreciate it if you could check whether it is a feasible and clear solution.

1  the  (not properly closed peer disconnected) connection (mrcpv2) contains  "channel_table",  which contains all the channels using the  connection 

2  each mrcp_control_channel_t  in "channel_table" will contain ( in its "obj")  the mrcp_channel_t, whose "obj" will contains the mrcp_session_t (also mrcp_server_sessiont_t).  that mrcp_session's "obj" will contains the sofia_session(mrcp_sofia_session_t).

Am I right on this.

So, for the problem connection, I could just get a list of different sofia_session associated with it.   If I just call " nua_bye(sofia_session ), the sofia task  will eventually call mrcp_sofia_on_state_change to terminate the session  through mrcp_sofia_event_callback. This might  sound ad-hoc.  Will it work?     If not, could I find a way to  send a "message" to  sofia_agent->nua.   Then I could  call "nua_bye" inside the mrcp_sofia_event_callback?

Anyway, I am going to try it out.

Appreciate it very much if you could enlighten me a bit.

Rgds,

[jerry]

Hi Arsen,

Seems it works. Thx!