Summary:ASTERISK-14333: Segfault on agent login, with manager, agi and pgsql realtime
Reporter:Fredrik Liljegren (fiddur)Labels:
Date Opened:2009-06-18 03:47:11Date Closed:2011-06-07 14:00:59
Versions:Frequency of
Environment:Attachments:( 0) gdb.txt
( 1) gdb-and-debug-2.txt
Description:Asterisk segfaults on agent login.  In our scenario, the following happens:

1. User calls 200
2. agi-script login.pl performs:
  1. takes an id through application Read.
  2. When login succeeds, it inserts into realtime table queue_member_table
  3. reads the variable QUEUE_MEMBER($queue) for each queue the agent is inserted into
  4. sends a UserEvent.
3. The UserEvent is catched by the manager:
  1. It pauses the queuemember via QueuePause.

The crash won't happen if the manager doesn't send the QueuePause command.

This works well in 1.6.0
The crash happens in svn trunk too.
Comments:By: Fredrik Liljegren (fiddur) 2009-06-18 04:34:03

I've flagged this for res_config_pgsql because it seems to be where that crash is... but it is triggered from (managers) QueuePause... I think.   With a changed login (that puts pause directly in psql and doesn't send AMI QueuePause), I get the crash on an unpausing from QueuePause with paused = false...

Trying to narrow it down further...

By: Fredrik Liljegren (fiddur) 2009-06-18 04:48:10

Ok, I can reproduce this crash from CLI with queue pause SIP/3011

Adding a new backtrace and debug-log.  What I did here was just:
CLI> queue pause SIP/3011

-> gdb-and-debug-2.txt

By: Matthew Nicholson (mnicholson) 2009-11-10 13:23:36.000-0600

You seem to be experiencing this crash when asterisk attempts to update a database field. I cannot reproduce it on my system using asterisk trunk r229102. Can you please specify which branch and revision you are using to test?

By: Fredrik Liljegren (fiddur) 2009-11-11 09:25:52.000-0600

It works well in trunk r229431 now, great!  I don't know what version I tried in, but i tested latest trunk a couple of times.  ...don't know why I didn't give the revision id on reporting...

By: Matthew Nicholson (mnicholson) 2009-11-11 09:33:02.000-0600

Ok, reopen this if you notice the crash in any other branches.  Thanks.