22 Aug
2011
22 Aug
'11
5:16 p.m.
Hello,
It seems kamailio 3 does not save message flags set in branch route into the
transaction. In reply_route and failure_route the flag set in branch_route is
unset. In 1.4 this used to work. I would like to get that behaviour back, is
this possible?
Testscenario:
route {
setflag(0);
t_on_branch("1");
t_on_reply("1");
t_on_failure("1");
xlog("L_NOTICE", "[$rm] Request before relay: $mF");
t_relay();
}
branch_route[1] {
xlog("L_NOTICE", "[$rm] Branch begin: $mF");
setflag(8);
xlog("L_NOTICE", "[$rm] Branch end: $mF");
}
onreply_route[1] {
xlog("L_NOTICE", "[$rm] Reply ($rs) begin: $mF");
setflag(4);
xlog("L_NOTICE", "[$rm] Reply end: $mF");
}
failure_route[1] {
xlog("L_NOTICE", "[$rm] Failure ($T_reply_code): $mF");
}
Log output on 3.x:
[INVITE] Request begin: 00000000
[INVITE] Request before relay: 00000001
[INVITE] Branch begin: 00000001
[INVITE] Branch end: 00000101
[INVITE] Reply (100) begin: 00000001
[INVITE] Reply end: 00000011
[INVITE] Failure (408): 00000011
Log output on 1.4:
[INVITE] Request begin: 00000000
[INVITE] Request before relay: 00000001
[INVITE] Branch begin: 00000001
[INVITE] Branch end: 00000101
[INVITE] Reply (100) begin: 00000101
[INVITE] Reply end: 00000111
[INVITE] Failure (408) begin: 00000111
--
Greetings,
Alex Hermann
22 Aug
22 Aug
5:43 p.m.
New subject: [regression] Message flags set in branch_route not saved into transaction
On Monday 22 August 2011, Alex Hermann wrote:
It seems kamailio 3 does not save message flags set in
branch route into
the transaction. In reply_route and failure_route the flag set in
branch_route is unset. In 1.4 this used to work. I would like to get that
behaviour back, is this possible?
This patch seems to restore old functionality. Would this be ok to commit?
diff --git a/modules/tm/t_fwd.c b/modules/tm/t_fwd.c
index 30558ab..756e4b4 100644
--- a/modules/tm/t_fwd.c
+++ b/modules/tm/t_fwd.c
@@ -1510,6 +1510,8 @@ int t_forward_nonack( struct cell *t, struct sip_msg*
p_msg ,
clear_branches();
setbflagsval(0, backup_bflags);
+ /* update flags, if changed in branch route */
+ t->uas.request->flags = p_msg->flags;
/* don't forget to clear all branches processed so far */
Testscenario:
route {
setflag(0);
t_on_branch("1");
t_on_reply("1");
t_on_failure("1");
xlog("L_NOTICE", "[$rm] Request before relay: $mF");
t_relay();
}
branch_route[1] {
xlog("L_NOTICE", "[$rm] Branch begin: $mF");
setflag(8);
xlog("L_NOTICE", "[$rm] Branch end: $mF");
}
onreply_route[1] {
xlog("L_NOTICE", "[$rm] Reply ($rs) begin: $mF");
setflag(4);
xlog("L_NOTICE", "[$rm] Reply end: $mF");
}
failure_route[1] {
xlog("L_NOTICE", "[$rm] Failure ($T_reply_code): $mF");
}
Log output on 3.x:
[INVITE] Request begin: 00000000
[INVITE] Request before relay: 00000001
[INVITE] Branch begin: 00000001
[INVITE] Branch end: 00000101
[INVITE] Reply (100) begin: 00000001
[INVITE] Reply end: 00000011
[INVITE] Failure (408): 00000011
Log output on 1.4:
[INVITE] Request begin: 00000000
[INVITE] Request before relay: 00000001
[INVITE] Branch begin: 00000001
[INVITE] Branch end: 00000101
[INVITE] Reply (100) begin: 00000101
[INVITE] Reply end: 00000111
[INVITE] Failure (408) begin: 00000111
--
Greetings,
Alex Hermann
7:35 p.m.
New subject: [sr-dev] [regression] Message flags set in branch_route not saved into transaction
On Monday 22 August 2011 16:43:43 Alex Hermann wrote:
On Monday 22 August 2011, Alex Hermann wrote:
Nevermind, i just found t_flush_flags(), which does exactly what i need.
--
Alex Hermann
It seems kamailio 3 does not save message flags
set in branch route into
the transaction. In reply_route and failure_route the flag set in
branch_route is unset. In 1.4 this used to work. I would like to get that
behaviour back, is this possible?
This patch seems to restore old functionality. Would this be ok to commit? 
25 Aug
25 Aug
1:49 p.m.
New subject: [sr-dev] [regression] Message flags set in branch_route not saved into transaction
Hello,
On 8/22/11 6:35 PM, Alex Hermann wrote:
On Monday 22 August 2011 16:43:43 Alex Hermann wrote:
was the auto-update of the flags in 1.x? That would have made
t_flush_flags() redundant, or maybe the auto-update was in branch route
and t_flush_flags() purpose was for failure_route. I want to be sure we
don't omit things from 1.x -- t_flush_flags() was also forgotten, added
later after a report from Juha.
Thanks,
Daniel
On Monday 22 August 2011, Alex Hermann wrote:
Nevermind, i just found t_flush_flags(), which does
exactly what i need. It seems kamailio 3 does not save message flags
set in branch route into
the transaction. In reply_route and failure_route the flag set in
branch_route is unset. In 1.4 this used to work. I would like to get that
behaviour back, is this possible?
This patch seems to restore old functionality.
Would this be ok to commit?
2:02 p.m.
New subject: [sr-dev] [regression] Message flags set in branch_route not saved into transaction
On Thursday 25 August 2011, Daniel-Constantin Mierla wrote:
Hello,
On 8/22/11 6:35 PM, Alex Hermann wrote:
In 1.4, message flags were updated in the transaction unconditionally after
branch_route had run. The patch i provided is an exact copy of the relevant
code from 1.4 and provides backward compatibility.
Adding t_flush_flags() to the end of branch_route in the script accomplishes
the same functionality, but requires a script update by the user.
IIRC, flags were also copied to the transaction from failure_route in 1.4.
Haven't tested though.
--
Greetings,
Alex Hermann
On Monday 22 August 2011 16:43:43 Alex Hermann
wrote:
was the auto-update of the flags in 1.x? That would have made
t_flush_flags() redundant, or maybe the auto-update was in branch route
and t_flush_flags() purpose was for failure_route. I want to be sure we
don't omit things from 1.x -- t_flush_flags() was also forgotten, added
later after a report from Juha. On Monday 22 August 2011, Alex Hermann wrote:
Nevermind, i just found t_flush_flags(), which does exactly what i need. It seems kamailio 3 does not save message flags
set in branch route
into the transaction. In reply_route and failure_route the flag set in
branch_route is unset. In 1.4 this used to work. I would like to get
that behaviour back, is this possible?
This patch seems to restore old functionality. Would this be ok to
commit?
2:07 p.m.
New subject: [sr-dev] [regression] Message flags set in branch_route not saved into transaction
On 8/25/11 1:02 PM, Alex Hermann wrote:
--
Daniel-Constantin Mierla -- http://www.asipto.com
Kamailio Advanced Training, Oct 10-13, Berlin: http://asipto.com/u/kat
http://linkedin.com/in/miconda -- http://twitter.com/miconda
On Thursday 25 August 2011, Daniel-Constantin Mierla
wrote:
I see, then we should add the
patch. Using t_flush_flags() will become
redundant for this case, but harmless.
Hello,
On 8/22/11 6:35 PM, Alex Hermann wrote:
In 1.4, message flags were updated in the
transaction unconditionally after
branch_route had run. The patch i provided is an exact copy of the relevant
code from 1.4 and provides backward compatibility. On Monday 22 August 2011 16:43:43 Alex Hermann
wrote:
was the auto-update of the flags in 1.x? That would have made
t_flush_flags() redundant, or maybe the auto-update was in branch route
and t_flush_flags() purpose was for failure_route. I want to be sure we
don't omit things from 1.x -- t_flush_flags() was also forgotten, added
later after a report from Juha. On Monday 22 August 2011, Alex Hermann wrote:
> It seems kamailio 3 does not save message flags set in branch route
> into the transaction. In reply_route and failure_route the flag set in
> branch_route is unset. In 1.4 this used to work. I would like to get
> that behaviour back, is this possible?
This patch seems to restore old functionality. Would this be ok to
commit?
Nevermind, i just found t_flush_flags(), which does exactly what i need.
Adding t_flush_flags() to the end of branch_route in the script accomplishes
the same functionality, but requires a script update by the user.
IIRC, flags were also copied to the transaction from failure_route in 1.4.
Haven't tested though.
We have to check and make it the same. The
t_flush_flags() will stay
then for the route blocks, where the transaction is created with
t_newtrans() -- which seems to be its description in 1.4.x:
http://www.kamailio.org/docs/modules/1.4.x/tm.html#id2531069
Cheers,
Daniel
4844
days inactive
4847
days old
5 comments
2 participants
participants (2)
-
Alex Hermann -
Daniel-Constantin Mierla