[Serusers] sl module functions in failure route

[Jiri Kuthan]

At 21:17 11/07/2007, Martin Hoffmann wrote:
>Jiri Kuthan wrote:
>> At 20:12 11/07/2007, JF wrote:
>> >So, the following snippet doesn't make sense in a failure route?
>> >I may try to call t_relay inside a failure route, and, if it fails,
>> >send a stateless reply back...
>> >
>> >       if (!t_relay()) {
>> >               sl_send_reply("500","Error relaying");
>> >               exit;
>> >       };
>> >
>> 
>> This piece of config makes sense, but it does not include failure_route
>> as you had asked before. From within failure_route, call t_reply to
>> complete transactino processing, don't call sl_send_reply.
>
>Wait. Does t_reply() skip branch picking? 

frankly, I'm not sure without looking in today's code. Anyway, it is the safe
way to generate a reply, otherwise transaction module will send what it considers
a good choice, which may be different from the code generated by sl_send_reply,
resulting on client confusion.

>If not, your t_reply("500",
>"Whatever") will most likely not get picked.
>
>What you need to remember in a failure route is that you already have a
>reply from older branches. If nothing came from downstream, SER
>fakes a 408. This means that you actually don't need to do the
>sl_send_reply() anyways. 

It is not that you don't need to, it is that you must not. Otherwise
there can be two replies, the SL-one and another one produced by TM
unaware of what SL has done.

>Just break out of the failure route and
>whatever reply lead to the call of the failure route will be send back
>upstream just fine. Which most likely is what you want anyways, if you
>think about it.


yes.

-jiri