Re: ch_malloc of 8388608 bytes failed (ITS#2270)

[joseph.tingiris@cox.net]

I have tested without kerberos and sasl compiled in and the problem remains.
It appears to  be a problem with replication because the master never
crashes at all .. only the replicas.  When the master isn't replicating, I
don't get the same malloc failures ... although, I'm still trying to track
it down.  How do I provide a stack back trace from the core dump?

I'm using the GNU linker when assembling, but I'm also using Solaris crle
and dynamic libraries.  I found this article and applied this patch with no
success.
http://sunsolve.sun.com/private-cgi/retrieve.pl?doc=patches%2F109147&zone_32
=malloc%20failed

I've also identified the following Solaris patch but haven't applied it yet.
http://sunsolve.sun.com/private-cgi/retrieve.pl?doc=fpatches%2F108827&zone_3
2=malloc%20failed  It looks promising.

Also, this is a multiprocessor 64bit system but I compiled and linked the
binaries with 32bit libraries.   I have a hunch this may be a long integer
problem or something.  What I find strange is that it is literally always
the same message but it affects different threads, of course.

slapd[237]: [ID 703374 local4.debug] ch_malloc of 8388608 bytes failed

Exactly 8 megabytes, every time.  Although, I can't seem to find out how to
reproduce it, yet.  I'm considering using dmalloc or something similar when
I compile to see if I can catch the malloc failure ... I'm in the process of
setting up a dedicated test environment just to solve this problem.

Thanks again for your help,

Joseph

----- Original Message -----
From: "Kurt Zeilenga" <openldap-its@OpenLDAP.org>
To: <joseph.tingiris@cox.net>
Sent: Wednesday, January 29, 2003 5:33 PM
Subject: Re: ch_malloc of 8388608 bytes failed (ITS#2270)


> Please provide a stack back trace from the core dump.
> Also, please test without MIT krb loaded as MIT krb is not thread-safe
> (use Heimdal instead).
>