[Home]

Summary:ASTERISK-10101: RFC2833 DTMF digits seem to be recognized, but the resulting sequence missing some digits
Reporter:mdu113 (mdu113)Labels:
Date Opened:2007-08-16 16:43:18Date Closed:2007-09-04 08:29:10
Priority:MinorRegression?No
Status:Closed/CompleteComponents:Core/Channels
Versions:Frequency of
Occurrence
Related
Issues:
Environment:Attachments:( 0) 10476-missing-digit-voicemail-log.txt
( 1) log.txt
Description:This is very strange situation. DTMF digits seem to be recognized, but the resulting sequence missing some digits.
Scenario: in VoiceMailMain i'm trying to dial a password. As you can see in uploaded log.txt all digits of the password '0501' were recognized, at least there's DTMF begin and DTMF end lines for them. Nonetheless it says
"-- Incorrect password '001' for user '13' (context = ace)"
It missed digit '5' when it checked the password.
This is intermittent, but usually I can replicate it once every 4/5 attempts.
This happens only with rfc2833 DTMF mode. Inband DTMF seems to be fine.
I'm not sure this is the right category I placed this report in as I'm not sure what it is. I apologize if this is placement is incorrect.


****** ADDITIONAL INFORMATION ******

One more detail. I've tested it on production and development server which are identical in terms of software and setup, but the production server is more modern and faster hardware. I can't replicate it on development server, but it happens on production which is much faster. So my wild guess it could be timing related somehow.
Please let me know if you need any more info.
Comments:By: Joshua C. Colp (jcolp) 2007-08-16 19:55:14

Please try the 1.4 branch in subversion. I made a change that should fix this.

By: mdu113 (mdu113) 2007-08-16 21:40:13

Hm, could you tell me what revision your fix is in?
At the moment there are 3 commits after 79690 and they seem to be unrelated.
The latest now is r79792

By: pj (pj) 2007-08-17 01:50:36

I have similar issue SVN-trunk-r79471, I'm using skinny phone (7920), when I enter WaitExten(), usualy only first dtmf dgit is recognized. This appears also before (I think it starts appearing after vldtmf feature was introduced), but before helps, when I typed digits slowly, now I'm not able to correctly send all digits at all.
same issue with all channels: skinny, sip, iax.



By: pj (pj) 2007-08-17 03:07:46

tried current SVN-trunk-r79824, no change:(

By: dtyoo (dtyoo) 2007-08-31 09:43:52

We are also experiencing something very similar to the problem reported by mdu113 with polycom ip 430 and 501 phones connecting to asterisk 1.4.9.  We recently upgraded from 1.2 to 1.4 and did not have this issue previously.  As best I can tell the DTMF (rfc2833) is making it all the way from the phone to the asterisk server, but then we see the voicemail application report "Incorrect password" with typically the first digit of either the mailbox or the password missing.  Switching to inband improves but does not eliminate this issue for us.  Unfortunately, the polycoms don't support sip info.

I'll post some of my trace / debug info in case anyone can get anything out of it.  The one other strange thing I have noticed a couple times is when this issue happens, my asterisk console session hangs for a moment, then comes back.

By: dtyoo (dtyoo) 2007-08-31 10:03:56

I just uploaded a snippet from an rtp debug session.  It shows the dtmf events for the mailbox (4011) and the password (2468), followed by:

   -- Incorrect password '2468' for user '011' (context = home)

It seems the 4 didn't make it to the voicemail application even though its in the rtp debug.

I'm happy to provide any further information needed or to do any further testing required.

By: Digium Subversion (svnbot) 2007-09-04 08:28:00

Repository: asterisk
Revision: 81437

------------------------------------------------------------------------
r81437 | file | 2007-09-04 08:28:00 -0500 (Tue, 04 Sep 2007) | 4 lines

(closes issue ASTERISK-10101)
Reported by: mdu113
Only look for the end of a digit when waiting for a digit. This in turn disables emulation in the core.

------------------------------------------------------------------------

By: Digium Subversion (svnbot) 2007-09-04 08:29:10

Repository: asterisk
Revision: 81438

------------------------------------------------------------------------
r81438 | file | 2007-09-04 08:29:10 -0500 (Tue, 04 Sep 2007) | 12 lines

Merged revisions 81437 via svnmerge from
https://origsvn.digium.com/svn/asterisk/branches/1.4

........
r81437 | file | 2007-09-04 10:46:23 -0300 (Tue, 04 Sep 2007) | 4 lines

(closes issue ASTERISK-10101)
Reported by: mdu113
Only look for the end of a digit when waiting for a digit. This in turn disables emulation in the core.

........

------------------------------------------------------------------------