Hi, I've finally managed to understand what was causing the crash. I've submitted a pull request with a proposal for a fix ( https://github.com/kamailio/kamailio/pull/346). Let's wait Daniel thoughts about it :)
Regards,
Federico
On Sun, Sep 20, 2015 at 4:17 PM, Federico Cabiddu < federico.cabiddu@gmail.com> wrote:
Hi Thibault, there is no need for special decline handling when using tsilo, so the default basic should work with the modifications you've done. I tested your scenario with this config and I haven't been able to get the crash. Could you please perform a test with debug level 3 and send me the whole log and the trace, sending them to my email address or sharing them?
Regards,
Federico
On Fri, Sep 18, 2015 at 3:38 PM, Thibault Gueslin < thibault.gueslin@gmail.com> wrote:
Hi Federico,
Thank you to keep looking at this.
I have sent you all my changes to default config. So there is nothing done after t_is_canceled (no voice mail, ...)
I have just enabled WITH_NAT & WITH_NATSIPPING.
I am sure if it is related but do you do special tricks for DECLINE (when enabling tsilo) ? (
The trace is always ending up with:
Program terminated with signal SIGSEGV, Segmentation fault.
#0 0x00007f5741b4efd7 in free_faked_req (faked_req=0x7f5741e00100 <faked_req>, t=0x7f573adb3c60) at t_reply.c:931
931 faked_req->body->free(&faked_req->body);
(gdb) p* faked_req->body->free
Cannot access memory at address 0x312e3838312e3832
It seems that the address of the "free" pointer is corrupted somewhere. I think I need to trace when it is modified (if it is possible).
Regards
thibault