[Home]

Summary:ASTERISK-07077: Segmentation fault on regular basis
Reporter:Todd Shore (teran)Labels:
Date Opened:2006-06-01 14:53:21Date Closed:2006-06-02 09:53:37
Priority:CriticalRegression?No
Status:Closed/CompleteComponents:Core/General
Versions:Frequency of
Occurrence
Related
Issues:
Environment:Attachments:( 0) btfull.txt
Description:Asterisk is crashing with a segmentation fault every few hours.  Problem appeared under 1.2.5 and continues with current version 1.2.7.1.  Runs fine when idle (weekend).  Crashes with varying call load.

This system does have Metermaids patch applied, though the system crashes in stock 1.2.5 and 1.2.7.1 form.
Comments:By: Serge Vecher (serge-v) 2006-06-01 15:02:16

does the crash happen when you blind-transfer a SIP call into Park()?

If above is not the case, you will need to produce a bt from non-optimized Asterisk, with metermaid patch removed. Also 1.2.8 is the latest version, with many many bugfixes big and small jstuk.

By: Todd Shore (teran) 2006-06-01 15:12:51

Users are parking calls using ASTERISK-694, which is assigned to DTMF key on their Snom phones.

The 1.2.7.1 that is running was compiled with dont-optimize.

With a fresh make of 1.2.7.1 (without Metermaids) I could get it to crash after about 10 parks.

I can move up to 1.2.8 tonight.

By: Serge Vecher (serge-v) 2006-06-01 15:17:11

Teran: this is a known issue in the current stable Asterisk. Please patch your 1.2.8 sources with a fix in 7053 and report the results.

Thanks.

By: Todd Shore (teran) 2006-06-01 15:24:52

I had read that issue but the fact that they are getting a hang instead of a crash and that they lose the ability to initiate calls while mine continue (until it crashes) made me think it was a different issue.

Anyway, I'll move to 1.2.8 tonight and see what happens.

By: Serge Vecher (serge-v) 2006-06-02 09:53:37

Teran: let's look for a solution in 7053 as I believe they are one and the same, albeit with slightly different symptoms. If the fix there does not resolve your case, please feel free to reopen the bug. Thanks.