Autor: Marc Schiffbauer Data: Para: exim-users Assunto: Re: [exim] Now exim segfault in a very strange way :-/ (was: Re: Why
does exim4 stat /boot and /proc/<pid>/mounts very often?)
* Philip Hazel schrieb am 17.12.04 um 15:14 Uhr: > On Fri, 17 Dec 2004, Marc Schiffbauer wrote:
>
> > sorry if I my question was a bit misleading. I know strace.. but I
> > wondered how to call exim with strace on a production box that is
> > already running the exim daemon to reproduce the stat() calls..
>
> Oh, I wasn't thinking of stracing a production system. You should be
> able to strace a single mail delivery without having to mess with the
> daemon.
ah ok. Will try that later.
>
> > Now I tried to send an email (using my local exim3 that connects to
> > the exim 4.34 that we are talking about). But that did not work
> > anymore! (after the strace call on the production box)
>
> > I asked a friend to try that telnet call.
> > Success. (?!!??) I use dialup for my internet connection
> > via some different providers... so I changed the provider... and
> > (STRANGE!) it worked now for me too... exim3 could empty its queue
> > and deliver the testmessage to my production box.
> >
> > Reverting back to the other provider restores the problem although I
> > got a different IP.
>
> That is, of course, completely weird. I'm afraid I've never used dialup
> and know nothing about it.
Does exim store hostnames and/or domains in any db file it uses? May
it be that there is a broken entry caused by the crash that makes
exim4 crashing when reading that entry again?
>
> > I logged one of that strace calls where it was segfaulting. Is that
> > of interest for you?
>
> If it is Exim 4 then yes, though I'm not too interested in 4.34 any
> more; if it is Exim 3 then I'm not interested, I'm afraid.
Its the debian woody backport of exim 4.34. And yes the trace is
from that exim. I will PM you the trace file, ok?
