I tried adding a watchpoint in GDB I do not seen when how the data is replaced by empty characters ...

On Fri, Jul 31, 2020 at 3:31 PM Julien Chavanton <jchavanton@gmail.com> wrote:

Sorry, the example was wrong ....

code :

void get_rpl_totag(struct cell *t, struct sip_msg *reply, str *totag) {
        if (reply==FAKED_REPLY || !reply || !reply->to) {
                tmb.t_get_reply_totag(t->uas.request, totag);
        }
        if (totag && totag->s) {
                LM_ERR("reply_totag ok [%p][%d][%*.s]\n", totag->s, totag->len, totag->len, totag->s);
        }
}

...

str ttag = STR_NULL;
get_rpl_totag(t, reply, &ttag);
LM_ERR(">>> ttag[%p][%d][%.*s]\n", ttag.s, ttag.len, ttag.len, ttag.s);

outputs :

0(25920) ERROR: acc [acc_logic.c:97]: get_rpl_totag(): reply_totag ok [0x7f0559c2a1e0][41][                                         ]
0(25920) ERROR: acc [acc_logic.c:525]: on_missed(): >>> ttag[0x7f0559c2a1e0][41][594d50c3218065a60bb91fd47a70fbc1-857d6855]

On Fri, Jul 31, 2020 at 3:26 PM Julien Chavanton <jchavanton@gmail.com> wrote:

I can not find the explanation on why the data memory is different once I return from the function.

This is the same process and the same memory address.

 

code :

void get_rpl_totag(struct cell *t, struct sip_msg *reply, str *totag) {
        if (reply==FAKED_REPLY || !reply || !reply->to) {
                tmb.t_get_reply_totag(t->uas.request, totag);
        }
        if (totag && totag->s) {
                LM_ERR("reply_totag ok [%p][%d][%*.s]\n", totag->s, totag->len, totag->len, totag->s);
        }
}

str ttag = STR_NULL;
get_rpl_totag(t, reply, &ttag);
LM_ERR(">>> to-tag[%p][%d][%.*s]\n", acc_env.to_tag.s, acc_env.to_tag.len, acc_env.to_tag.len, acc_env.to_tag.s);

outputs :

 0(25216) ERROR: acc [acc_logic.c:97]: get_rpl_totag(): reply_totag ok [0x7fa6a79d81e0][41][                                         ]
 0(25216) ERROR: acc [acc_logic.c:525]: on_missed(): >>>>>>>>>>> to-tag[0x7fa6a79d81e0][41][594d50c3218065a60bb91fd47a70fbc1-857d6855]

On Fri, Jul 31, 2020 at 8:33 AM Julien Chavanton <jchavanton@gmail.com> wrote:

Seems to work, I will complete the modifications and make an MR ...

On Fri, Jul 31, 2020 at 12:30 AM Julien Chavanton <jchavanton@gmail.com> wrote:

I can see TM API is exposing

tmb.t_get_reply_totag(msg, &to_tag);

looking at what can be done from acc_logic.c

on_missed

/* set env variables */
env_set_to( get_rpl_to(t,reply) );
env_set_code_status( code, reply);
...

On Thu, Jul 30, 2020 at 11:51 PM Julien Chavanton <jchavanton@gmail.com> wrote:

Hi, considering this example :


modparam("acc_json", "acc_missed_flag", 6)

failure_route[test] {
    setflag(6); // MISSED_ACC, the failure will be recorded by ACC
    t_flush_flags(); // This will set the flags in the newly created transaction
    t_reply("500", "Service Unavailable");
}

This does generate an ACC event, however the to-tag is not populated, my guess this is because the TM callback is done before the to-tag is generated.

I wonder if there is a way to populate the to-tag immediately ?

I will dig it further, but maybe there is a trick I am missing