[libnice] Assert due to wrong transition

Igor Khomenko igor at quickblox.com
Fri Mar 23 16:40:14 UTC 2018


Thanks Fabrice for getting back to me

is there a custom Git branch with the patch I can build from or should I
port all those changes manually?

At this moment I use latest master -  commits on Nov 28, 2017

thanks


On 23 March 2018 at 18:32, Fabrice Bellet <fabrice at bellet.info> wrote:

> Hi Igor,
>
> On 03/22/18 at 10:12am, Igor Khomenko wrote:
> > Hi Olivier, Lorenzo
> >
> > I've got all needed logs upon crash, including libnice logs
> >
> > Stacktrace again (the same assert):
> > https://www.dropbox.com/s/jlqtzu0neqlqc8o/janus_libnice_
> assert_stacktrace.crash?dl=0
> >
> > logs:
> > https://www.dropbox.com/s/hq1nkeh6w25jsk3/janus.log?dl=0
> >
> > look at the end of janus.log
> > Line N19365 - "(process:17784): libnice-DEBUG: Agent 0x7f6aa000dcf0 :
> > stream 1 component 1 STATE-CHANGE gathering -> connected."
> > this is when crash happened, then autorestart of Janus occurred after
> some
> > time.
> >
> > So all the lines before N19365 are what we are looking for from my
> > understanding,
> > e.g. line N19330 is where assertion failed
> >
> > Please let me know if you need anything else here to identify the root
> > cause and find a proper solution
>
> According to your logs, I think the cause of the problem is that no pair
> is formed by the gathering stage: we have two local IPv4 candidates, and
> we receive a remote IPv6 candidate. A new pair is formed later when we
> start receiving incoming STUN requests. The state of the agent should
> change from GATHERING to CONNECTING there, line 19285 in your janus.log
> file:
>
> (process:17784): libnice-DEBUG: Agent 0x7f6aa000dcf0 : Adding a triggered
> check to conn.check list (local=0x7f6aa00239f0).
> (process:17784): libnice-DEBUG: Agent 0x7f6aa000dcf0 : creating new pair
> 0x7f6aa001a190 state 1
>
> Does this patch help in your case : https://phabricator.
> freedesktop.org/D1889
> And more precisely, the chunk around line 2967-2968 in conncheck.c that
> replaces
> priv_add_new_check_pair() by priv_conn_check_add_for_
> candidate_pair_matched()
> at the end of function priv_schedule_triggered_check()  ?
>
> Best wishes,
> --
> fabrice
>



-- 
Igor Khomenko,
CTO @ *QuickBlox*

igor at quickblox.com <taras at quickblox.com>
Q-municate <https://q-municate.com/>: Igor Khomenko

We will be happy to solve your tech issues if you ask ‘quickblox’ tagged
question on StackOverflow http://stackoverflow.com/questions/ask
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/nice/attachments/20180323/8a350d35/attachment.html>


More information about the nice mailing list