Summary:ASTERISK-14253: [patch] segfault (sig 6)
Reporter:caspy (caspy)Labels:
Date Opened:2009-06-02 11:14:17Date Closed:2011-06-07 14:00:36
Versions:Frequency of
Environment:Attachments:( 0) 20090325__debug_bad_memory_without_valgrind.diff.txt
( 1) bt_20090602_1947.txt
( 2) bt_20090603_1055.txt
Description:died with signal 6, i can't understand where and why.
backtrace attached.
Comments:By: caspy (caspy) 2009-06-02 11:15:31

also, no one unusual messages in logs at all (even in debug w/level 1)

By: Russell Bryant (russell) 2009-06-02 11:17:18

Can you try with DONT_OPTIMIZE?  The current backtrace is useless ...

By: Tilghman Lesher (tilghman) 2009-06-02 11:39:37

Actually, the backtrace tells me that you'll need to follow the instructions in doc/valgrind.txt, as this is likely a memory corruption error.

By: caspy (caspy) 2009-06-02 12:07:29

russell, it is already DONT_OPTIMIZE. 100%. look for my other open issues.

tilghman, valgrind is not an option in my case. it is a loaded production, and valgrind's slowness is unaccepable for me. :(

By: Tilghman Lesher (tilghman) 2009-06-03 00:02:26

Okay, here is a debugging patch that provides us with similar information to valgrind, without the corresponding slowdown.  Patch, then run 'make menuselect', and select the option MALLOC_HOLD under Compiler Options.

By: caspy (caspy) 2009-06-03 02:05:33

with this patch can't even start.
backtrace attached: bt_20090603_1055.txt

By: caspy (caspy) 2009-06-11 04:24:13

any ideas?

By: Leif Madsen (lmadsen) 2009-07-13 10:21:00

Setting back to status New.

By: Tilghman Lesher (tilghman) 2009-08-25 17:11:26

You could try running the /team/tilghman/malloc_hold branch (run 'make menuselect' and under Compiler Options, select MALLOC_HOLD before compiling).

By: caspy (caspy) 2009-08-26 08:34:06

there was no such crashes since that time.
let's close bug for now, cause we can wait years until it crash again.

if it will come back - i'll write here to reopen bug.

By: Tilghman Lesher (tilghman) 2009-08-26 10:06:51

Actually, given that each crash like this cannot be conclusively linked, other than the fact that signal 6 killed it, we'd prefer if you opened a new issue.