Re: [Bodington-developers] More LDAP Fun and Games

[Jon M. <jo...@te...>]

Sean Mehan wrote:
> cool, let's see some ping and bing times to quantify this network  
> congestion difference.
>   
The fastest link was via the localhost device and when I tried against 
the University AD it was a dedicated 100Mbs ethernet line to a router on 
the really really fast ATM network (I think). The slower connection that 
works is through my high latency 54Mbps wireless network.
>
> what are the kernels and distros?
>   
PC That doesn't throw exceptions;
Linux version 2.6.9-1.667 (bhc...@tw...) (gcc 
version 3.4.2 20041017 (Red Hat 3.4.2-6.fc3)) #1 Tue Nov 2 14:41:25 EST 2004
PC That does;
Linux version 2.6.17-1.2145_FC5 
(bre...@hs...) (gcc version 4.1.1 2006 0525 
(Red Hat 4.1.1-1)) #1 Sat Jul 1 13:03:45 EDT 2006

I've been digging in the source code for OpenLDAP and I think it's 
related to poor coding in relation to handling timings.  As a test of 
that theory I set the LDAP search timeout to infinity and the exceptions 
stopped. Setting the search timeout to any finite value causes 
exceptions on one particular client PC.  Of course I also chopped out 
the custom Socket Factory but that made no difference at all.  My 
sneaking suspicion is that the conversation with the LDAP server starts 
in one thread while another thread prepares to receive the response but 
the response comes back before that setup is complete. Mostly it seems 
to be trying to clear semaphores that don't exist and carries on 
executing anyway but from time to time it just locks up.