[tpop3d-discuss] signal 11?

[Nancy Pettigrew (work)]

Well, the server died again today, but this time we got a core file (the
parent process died too).  Did the backtrace and found that the y variable
in auth_mysql_new_user_pass was out of bounds.  Now, I'm initializing the
vars at the start of the function.

Thanks for your suggestions!

-----Original Message-----
From: tpop3d-discuss-admin@lists.beasts.org
[mailto:tpop3d-discuss-admin@lists.beasts.org]On Behalf Of Paul
Makepeace
Sent: Tuesday, July 17, 2001 5:24 PM
To: Nancy Pettigrew (work)
Cc: tpop3d-discuss@lists.beasts.org
Subject: Re: [tpop3d-discuss] signal 11?


On Tue, Jul 17, 2001 at 05:14:34PM -0700, Nancy Pettigrew (work) wrote:
> Problem for me is that there is no core file to be found and the error is
> sporadic;

Have you tried setting ulimit -c unlimited (as below) so your
environment permits writing core files?

> so far I've been unable to recreate it in a test environment.

I am currently working on integrating tpop3d, exim and Oracle and part
of the requirements for deployment is a documented performance test.
Oracle comes with its own diagnostics and benchmark system but it would
be good to have a configurable expect-based tool to concurrently test
tpop3d (and exim). With the dial cranked up it should be possible to
flush out bugs, race conditions, etc by beating it increasingly hard.

Has anyone done anything like this?

Paul

--snip--

I wrote:
> One thing that might help you is to do a backtrace as above. Invoke "gdb
> tpop3d core" and type "backtrace" or "bt". Ensure you're in a position
> to write core files with (with bash at least) ulimit -c unlimited.

--
Shut the door and listen from outside

_______________________________________________
tpop3d-discuss mailing list
tpop3d-discuss@lists.beasts.org
http://lists.beasts.org/mailman/listinfo/tpop3d-discuss