[sr-dev] DMQ: seems broken logic

[Daniel-Constantin Mierla]

Hello,

On 15/10/14 20:36, Charles Chance wrote:
> Hi Daniel,
>
> On 15 October 2014 14:51, Daniel-Constantin Mierla <miconda at gmail.com
> <mailto:miconda at gmail.com>> wrote:
>
>     Hello,
>
>     one thing I wanted to add -- it may be good to backup the old
>     value of the sock and restore it later.
>
>
> Thanks - I had that in mind already.
>  
>
>     I am not sure if this is used in a context of a request that is
>     going to be forwarded as well to some other address. If yes, then
>     perhaps is it safer to backup and restore initial socket value.
>
>     Perhaps it needs to check if the transaction exists and if not,
>     create it first via tm api, then set the new socket for message
>     and call the t_replicate function, not to get the dmq server
>     socket in tm.
>
>
> Sorry, I don't fully understand the last part?
t_replicate() create the transaction for the respective requests (if the
transaction doesn't exist already). Creation of the transaction involves
cloning to shared memory the sip_msg_t structure. If you update the
forced socket to dmq server address for the request, it gets in shared
memory with the new value, which may be wrong for further processing of
the message in branch/failure routes.

My suggestion is to check if transaction exists and if not, first create
it and then backup socket, update it for dmq address, do the replication
and restore the socket from backup (note that the dmq the function will
still use the sip_msg_t from private memory even after creating the
transaction).

Cheers,
Daniel

-- 
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sip-router.org/pipermail/sr-dev/attachments/20141015/8a3bd5b8/attachment.html>