Summary:ASTERISK-04768: One way audio betwen 1.0 and HEAD with trunktimestamps=no
Reporter:Danilo Lotina F. (dlotina)Labels:
Date Opened:2005-08-03 13:03:28Date Closed:2008-01-15 15:43:53.000-0600
Versions:Frequency of
Environment:Attachments:( 0) trunktimestamps.patch
Description:1.0 box runnig 1.0.9
HEAD with has trunktimestamps=no

call from HEAD to 1.0 box. No audio behind 1.0 box

Many WARNINGS on 1.0 box :
Aug  3 13:58:05 WARNING[9303]: chan_iax2.c:5067 socket_read: meta trunk cmd 1 received, I only understand 0 (perhaps the remote side is sending trunk timestamps?)

Comments:By: Mark Spencer (markster) 2005-08-03 13:52:30

I'm guessing you didn't remember to make the change on the "head" side, you only made it on the "stable" side?

By: Danilo Lotina F. (dlotina) 2005-08-03 14:14:14

Hi Mark:

Do you mean trunktimestampso change? that was made at head side.

call from A to B:

(A) --ZAP-->  [ * HEAD ] --IAX--> [ * 1.0 ] --ZAP--> (B)

* HEAD : Last CVS iax.conf has trunktimestamps=no
* 1.0  : 1.0.9
A hear B
B doesn't hear A

as workaround audio works two way if I disable trunk on peer.

By: Mark Spencer (markster) 2005-08-03 14:17:34

And can you please confirm this is still an issue with latest CVS head?

By: Danilo Lotina F. (dlotina) 2005-08-03 14:24:24

Yes,  HEAD box has cvs version from 1 hour ago.

By: Mark Spencer (markster) 2005-08-03 15:55:10

Find me on irc so I can login to the machine with head.

By: Russell Bryant (russell) 2005-08-04 16:33:15

Where do you have the trunktimestamps option set?

From looking at the code, even though it is a global option in the code, it is only read from a peer entry.

Try setting the option inside of your peer entry if that is not where you have it set currently.

By: Russell Bryant (russell) 2005-08-04 18:07:35

I have uploaded a patch that fixes the trunktimestamp option to be set in the general section instead of a peer section.  Previously, you had to set it in a peer definition, even though it sets the option globally.  

If someone would like to fix it up so you can set it on a per-peer basis as well, that would probably be a better way to do it, but I don't think it's that big of a deal.

This patch also includes a couple cleanup thingies.

1) The global trunktimestamps option has been changed to be a bit flag instead of using an int.

2) I have removed a variable 'found' from build_user that wasn't being used for anything.

By: Danilo Lotina F. (dlotina) 2005-08-04 20:49:48

drumkilla: You are right, trunktimestamps=no works only as peer definition.
I will test your patch.

By: Danilo Lotina F. (dlotina) 2005-08-05 23:22:34

patch works fine.

By: Mark Spencer (markster) 2005-08-06 10:49:16

Fixed in CVS head...  This option doesn't necessarily make a lot of sense to have as a per-peer option at the moment certainly but it might in the future.  I also changed default behavior to match that of CVS stable (i.e. do not by default include timestamps)...  Having them on by default was definitely a mistake I didn't catch in the original patch.

By: Russell Bryant (russell) 2005-08-06 11:20:08

Well, I figured you may have one peer running cvs head that you want to send timestamps to, and another running 1.0 where you do not.  *shrugs*

By: Digium Subversion (svnbot) 2008-01-15 15:43:53.000-0600

Repository: asterisk
Revision: 6293

U   trunk/channels/chan_iax2.c

r6293 | markster | 2008-01-15 15:43:52 -0600 (Tue, 15 Jan 2008) | 2 lines

Fix timestamp + trunking (bug ASTERISK-4768)