Re: [Exim] exim 3.22 problem on solaris 2.6

Sheldon Hearn wrote in message ID
<6349.991732161@???>:
>
>
> On Tue, 05 Jun 2001 09:43:53 +0100, EXIM Users wrote:
>
> > We are seeing LOTS of these appearing,
> > at first we thought it might be a memory/disk corruption issue but after
> > moving the bulk of our mail over to another solaris 2.6 server we started
> > seeing exactly the same problem.
>
> What compiler version did you use to build Exim and what compiler
> optimizations did you specify? If you used anything more aggressive
> than "-O", try reducing the optimization to "-O".
>
> I don't know for a fact that this will do anything useful, but it's
> always best to rule the compiler out first, and some of gcc's
> optimizations (e.g. "-O2") are known to produce subtle problems.

We always compile with -O after many bad experiences with GCC
optimizer bugs. Just FYI, we ran exim 3.22 with -O optimization on
Solaris 2.6 5/98 with kernel 105181-19, using gcc 2.95.2 and ran into
the problem. It seems it may be related to either the format of the
message or the sending host, as our inbound MX's did not experience
this problem evenly.

(We have a load balancing solution that selects a backend server based
on the source IP address, but all 5 machines are roughly evenly
loaded. However, a couple of the machines accumulated dead -D files
faster than the others)

Is this just happening on Solaris? Anyone seen it on other platforms?
If its just Solaris, it may be a weird library interaction.

Also, anyone ever get panic logs generated? I was under the
impression that exim checked the return from any children (including
itself) and would log any abnormal exits. We never saw any panic logs
generated, so I'm wondering if the panic log stuff is broken, or if
there is some situation where the exim child is left parentless and
gets inherited by init, so logs wouldn't be generated.

Thanks,

Gary