Summary:ASTERISK-07577: Asterisk 1.2.10 Crash - the MixMonitor issue
Reporter:Steve Davies . (stevedavies)Labels:
Date Opened:2006-08-22 04:36:43Date Closed:2006-09-18 10:21:46
Versions:Frequency of
Environment:Attachments:( 0) btfull.txt
Description:Asterisk 1.2.10 box, used for call logging - 2 PRIs in, 2 out.
MixMonitor is being used for the call recordings
Box ran stable on 1.2.9, trying to use 1.2.10 it crashes out every 10 minutes or so,
Glibc logs an error to STDOUT/STDERR reporting Memory Corruption, and (presumably deliberately) SIGABORTS.

Reverting the dialplan to use Monitor instead of MixMonitor does eliminate the problem.

I've opened a new bug - sorry, I know its a dupe of other reports, but perhaps this is a fresh start because (tada!) I have a bt full made from a dont-optimize build...

I have 7 core files, but the crash is the same in all of them.


There appear to have been 2 commits into app_mixmonitor.c between 1.2.9 and 1.2.10:

r33841 | kpfleming | 2006-06-13 15:30:06 +0200 (Tue, 13 Jun 2006) | 2 lines
memory allocation optimizations
r33724 | file | 2006-06-12 23:34:38 +0200 (Mon, 12 Jun 2006) | 2 lines
Greatly simply the mixmonitor thread, and move channel reference directly to spy structure so that the core can modify it.

Disclaimer on file (but no code in this report anyway!)
Comments:By: Eduardo Gallo (edgallo) 2006-08-31 21:46:03

Ok ... been there, done that. Just spent a whole week figuring it out, and I confirm stevedavies' findings. Tried to upgrade to 1.2.11 .. no good. Reverted to 1.2.9 ... is now working like a charm. Have already gone through the new code on both branch1.2 and trunk (as of tonight), and they seem fishy still.

Sorry guys .. I need my production mixmonitoring, will stick to 1.2.9 a little longer. :)

By: Joshua C. Colp (jcolp) 2006-09-05 15:06:18

I just backported the trunk fixes I did this weekend into 1.2 - please update to latest 1.2 branch and give it a try!

By: Joshua C. Colp (jcolp) 2006-09-18 10:21:45

No responses people, but that's okay. If an issue still exists please reopen this bug.