Another possibility is that exim is somewhere freeing a pointer twice,
though why this would should up only under load I don't know. I do know
that libcs after 5.2.18 are using a new malloc which is very intolerent
of this. (see one of the current faq's on netscape with the new libc)
Why don't you try starting exim from a shell script which sets
LD_PRELOAD=/lib/libc.so.5.2.18 or sets MALLOC_CHECK_=1 and see if the
problem goes away. I know I'm definately not seeing problems like this
with Solaris, though I'm only at 1.58 there.
If the above fixes the problem it's probably time to compile exim with
electric fence or one of the other malloc debuggers and see what's up.
On Apr 28, Christoph Lameter <clameter@???> wrote:
> I have compiled a kernel on it. If memory would be the issue then other
> applications than exim should segfault. They dont and I have this behavior
> on multiple machines.
>
> > Perhaps a high-turnaround Exim installation might stress a platform
> > sufficiently to bring out the same kind of SIGSEV symptoms as gcc can
> > trigger. Just a thought, and easy to verify: merely compile a kernel
> > on the platform in question, and if the compile doesn't fail then memory
> > flakiness is not the cause of the weird Exim behaviour.
>
> --- +++ --- +++ --- +++ --- +++ --- +++ --- +++ --- +++ ---
> Please always CC me when replying to posts on mailing lists.
>
--
Artificial Intelligence stands no chance against Natural Stupidity.
GAT d- -p+(--) c++++ l++ u++ t- m--- W--- !v
b+++ e* s-/+ n-(?) h++ f+g+ w+++ y*
