On 09/24/14 09:16, Sebastian Damm wrote:

Hi,

I switched from rtpproxy module to the rtpproxy-ng module lately, and noticed a strange behavior. In my branch route to the device, I have two statements:

fix_nated_sdp("1"); rtpproxy_offer();

The first command appends a line with "direction:active" to the SDP. The second one puts the RTP proxy in the stream. This worked all well with the old rtpproxy module. But with the new rtpproxy-ng module, I get an empty line after the SDP body, just before the "direction:active" line in the SDP, which makes the packet invalid.

I tried changing the order of both statements, but without any difference. What I saw is, that the old module just sent some basic parameters to the rtpproxy and got only IP and port back. The new module sends the complete SDP to the rtpengine and gets back the "fixed" SDP.

Has anyone ever seen this? Is there a way to fix it?

This is caused by rewriting the SDP body twice. The rtpproxy module only rewrites small parts of the SDP, while rtpproxy-ng substitutes the entire SDP body, which is why you run into conflicts when there's another function touching the SDP body. Calling msg_apply_changes() between the two calls fixes that, but may have side effects. I would suggest the same as Juha did.

cheers