On 30/06/14 12:04 PM, Alex Balashov wrote:
On 06/30/2014 09:48 AM, Richard Fuchs wrote:
After carefully considering the code again, I
believe the assertion
trigger may be a rare but valid case after all. I have thus removed the
assertion and the code is now simply ignoring this case. I expect that
fixes your problem.
https://github.com/sipwise/rtpengine/commit/2e9ae75fb0eeddaa877f27c9c4ba98a…
Thank you, Richard. Can you elaborate as to what this case might be?
It's about how many ports are used for a particular media stream,
normally it's two, but it can be different. If there's a disagreement
about this between the two sides, then I believe some of the pointers
could be left NULLed. I haven't actually tested this though.
There's also the issue of a race condition due to incorrect locking.
I've just pushed another commit to make this more robust. Either case
requires reception of a stray UDP packet on a local port which is either
unused or not yet used.
cheers