Re: (ITS#4667) assert core in connection.c

[quanah@stanford.edu]

--On Friday, September 08, 2006 8:41 PM +0200 Pierangelo Masarati 
<ando@sys-net.it> wrote:

> Quanah,
>
> I'll try to look at it.  Let me just add that I recently saw yet another
> core in that area; I didn't post about it because I was shutting down a
> hacked version of 2.3 with some HEAD and some custom code in it all
> interspersed.  It triggered an abort() after asserting on
>
>         assert( c->c_struct_state == SLAP_C_USED );
>
>
> in connection_state_closing(); I couldn't reproduce it.  Since things
> like this seem to occur mostly to you and me when using non-standard
> code, chances are we both don't use exactly all the necessary patches; in
> any case, there might be something subtle in that portion of code.

Well, this code is running on all my servers, and they aren't hitting it. 
I was only able to trigger it when putting OpenLDAP under extremely high 
load, and adding SASL/GSSAPI into the mix.  If I am missing patches, it 
isn't in connection.c or daemon.c.  Not sure where else to look past that. 
;)  I'll see if I'm able to reliably reproduce it or not.  If not, I'd 
guess it is a race condition of some sort somewhere...

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html