Closed GoogleCodeExporter closed 8 years ago
Here is the wireshark capture.
Original comment by vsoffice...@gmail.com
on 17 Dec 2009 at 8:33
Attachments:
The above Wireshark capture has Asterisk and UniMRCP running in one server and
softphone/Nuance Server running in another server (the capture is from the
Nuance
server). Here is yet another Wireshark capture that has only Nuance server
running
on it.
Original comment by vsoffice...@gmail.com
on 17 Dec 2009 at 9:29
Attachments:
Thanks for the network captures. I've looked through them and indeed SIP and
MRCP
message flow looks good. Moreover, I can clearly listen to synthesized speech
from
Nuance to Asterisk. Though, it seems nothing was sent from Asterisk to X-Lite.
So I continue to think this is configuration related issue. Can you please
provide
also mrcp.conf (the configuration of connector bridge) and probably Asterisk's
sip.conf.
Finally Asterisk console output with enabled UniMRCP logs would be very helpful.
I'll try to dig into closer and probably reproduce the same on my side tomorrow.
Original comment by achalo...@gmail.com
on 17 Dec 2009 at 9:49
Hi,
Here is the attachment from the Nuance server wireshark capture (192.168.1.100)
and also conf files and asterisk server logs (192.168.1.105). SIP Phone was at
192.168.1.103. I disallowed all except ulaw in sip configuration as you had
suggested.
Thanks,
VS
Original comment by vsoffice...@gmail.com
on 21 Dec 2009 at 2:19
Attachments:
Thanks for the detailed report. It seems I've started to understand what is
going on.
I've noticed no RTP packets were received from Nuance MRCP server by UniMRCP
module
plugged into Asterisk.
[Dec 20 06:05:30] NOTICE[3628] app_unimrcp.c: Close RTP Receiver
192.168.1.105:4002
<- 192.168.1.100:7892 [r:0 l:0 j:0]
However, UniMRCP client stack successfully bound to this port. Moreover, network
capture shows that packets were indeed sent to the proper network address such
as
192.168.1.105:4002.
So I know no good reason why packets can't be received from user (application)
space,
unless it's firewall or selinux related issue.
Please either explicitly open ports for RTP port range (4000 - 5000 by default)
or
probably better at least temporary disable firewall and selinux.
Original comment by achalo...@gmail.com
on 21 Dec 2009 at 1:21
Hi Guys,
this issue looks similar like the one I told Arsen about via e-mail few days
ago.
Older release was working fine (twinkle and linphonec softphones plus GVP
client),
but newer one not with twinkle.
As I said, I will do the new testing after the New Year.
Cheers
- Vali
Original comment by tomas.valenta@speechtech.cz
on 21 Dec 2009 at 1:26
Hi Vali,
Thanks for the update. I guess the issue you experienced could be related to
some
inaccuracy in SDP offer/answer assuming particular setup you have. Anyway, I'd
need
logs and traces to better identify who is guilty there. Please provide it when
possible.
In the meantime, very similar (almost the same) setup described in the scope of
this
issue is one to one working to me. So hopefully we are close to identify the
reasons.
Original comment by achalo...@gmail.com
on 21 Dec 2009 at 2:04
Hi there. Not sure I fully understand all of this, but it is interesting that
you use
a payload type of 97 for LPCM at 8000 samples/s. I use L16/96/8000 with my
Nuance
Speech Server 5.1.0, Recognizer 9.0.3 and Vocalizer 5.0.3 setups and this works
very
nicely. Perhaps this is why you get "unknown codec"/"incorrect samples" errors?
Also, in the code of app_unimrcp.c, you will see:
speech_channel_create_mpf_termination
initialises the payload to 96 for L16, 8000 in the case of UniMRCP versions
before 0.8.0.
Hope this may help!
Original comment by thirion...@gmail.com
on 22 Dec 2009 at 11:33
Hi Derik,
Great you're back to the project and thanks for the note.
I also initially thought the problem could be in different payload types in
offer/answer, though normally it's still should cause no problem.
However, if you check messages.txt from comment-4, you'll see that no RTP
packets
were received in either of calls/sessions. In the meantime,
AstNuanceServer12202009.pcap capture(from the same comment-4) shows, that
packets
were sent, though you may notice Asterisk sent ICMP packets back and RTCP
statistical
reports contained 0 received and this is the actual problem.
By default, all the recent Linux distributions including RHEL5 come with
Firewall and
SELinux enabled and I assume this may be the reason, unless I'm missing
something
very obvious here. So probably we should just wait a bit to get the answers.
Original comment by achalo...@gmail.com
on 22 Dec 2009 at 3:49
Hi,
The problem seems to be with the SE Linux settings. I had it in my bash profile
and Asterisk was not using this shell. The RTP packets were dropped due to this
setting. I have corrected it. It works fine now.
Original comment by vsoffice...@gmail.com
on 11 Jan 2010 at 10:02
Thanks for the update.
I think this might be a typical issue, so the final resolution should be
helpful for
others too.
Original comment by achalo...@gmail.com
on 12 Jan 2010 at 10:51
Original issue reported on code.google.com by
vsoffice...@gmail.com
on 17 Dec 2009 at 7:31