On Thu, Sep 01, 2016 at 03:07:46PM +0200, Daniel-Constantin Mierla wrote:
I pushed a new commit, aiming to avoid overwriting
record route set with
session updates. Let me know if it makes the difference.
The same scenario (callee wants sessiontimers refreshed by the uac)
works kind of. ACKs are routed correctly but:
-Route to the callee has trailing nulls (starting from the first ACK to
the callee (packet 7 in pcap))
-At random (seen it happen in both testcalls) the Via header to the
caller is borked (packet 77) in the 200 OK to the caller. Someone (not
the callee for sure, and doesn't look like a packet from the caller) is
generating a CANCEL (packet 95) and next the caller will hangup.
-When enabling topos module, kamailio leaks memory like crazy at a rate
of about 1 GB in 2 hours without any calls (only OPTIONS and replies)
BTW might any of these be because of the way I test? I update the git
tree and copy the topos module stuff into my 4.2 tree (debian source
package), build the package, and copy only topos.so to live.