[SR-Users] Asynchronous processing increases branch count

[Daniel-Constantin Mierla]

Hello,

On 11/12/15 02:33, Alex Balashov wrote:
> Hi,
>
> Normally with serial forking, the first branch ID corresponds to
> $T_branch_idx == 0, i.e.
>
> Via: SIP/2.0/UDP
> 172.30.110.4;branch=z9hG4bK7f3d.57593ee3a139f0ae6c4a8a0f531e342c.0
>
> In the event of failure, a second branch attempt has a branch index of
> 1 and so on:
>
> Via: SIP/2.0/UDP
> 172.30.110.4;branch=z9hG4bK7f3d.57593ee3a139f0ae6c4a8a0f531e342c.1
>
> However, when I use asynchronous processing with t_suspend() /
> t_continue(), resuming the transaction in an rtimer process, I notice
> that the $T_branch_idx of the first branch is 1, then 2.
>
> Where did 0 go? Is this a normal effect of TM suspend/continue?
yes, this is the normal. Suspending a transaction means using a branch
which is not sent out, just the timer is started for it (in case is not
resumed during retransmission timeout interval, it will be replied with
408 (or failure route executed), to avoid long lasting/memory leak by
improper config suspend-resume logic).

At some point in time there was an attempt to reuse the suspended
branch, but it didn't prove to be reliable, as some internal
flags/attributes could get overwritten there.

Cheers,
Daniel

-- 
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Book: SIP Routing With Kamailio - http://www.asipto.com
http://miconda.eu