[Devel] Race conditions in TM, SER implementation, etc...

[Jerome Martin]

Hi list,

I'm starting this as a new topic, because I'd really want to be able to
assess the risk I'm taking by using OpenSER vs SEr (I don't intend to
move to SER right now, however it would be good knowing exactly the
weaknesses  of OpenSER, as I'm using it in production right now).

What I'm interested in is understanding something Jiri is explaining
about race conditions in tm, about that probleme being fixed in last SER
cvs but still present in OpenSER 1.2 despite all the work by Bogdan on
timers etc...

Firsthand : Jiri, is that right or did I misunderstood what you've
said ?

Looking at both SER cvs and OpenSER, I see comments pointing to a
possible race condition here and there, typically in timer.c. 
So could anyone point me to the "risky operations" on timers that would
lead to race conditions with noticeable side-effects that still exist in
OpenSER but are solved in SER ?
What would those side effects be ? 
Is that a locking-related issue, or is it a deeper structural problem ?

Bogdan, what do you think ?

Best Regards,
Jerome





-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://openser.org/pipermail/devel/attachments/20070330/6f58798b/attachment.htm