Yury Burkatovsky <tby(_at_)telrad(_dot_)co(_dot_)il> writes:
Bjorn Wiren <etxbwir(_at_)cbe(_dot_)ericsson(_dot_)se> wrote on Thu, 02 May
2002:
I reported this error a week ago
(http://lists.ccil.org/pipermail/fetchmail-friends/2002-April/002035.html),
but so far there has been no response.
Have you been able to see if fetchmail detects some error before
dying? You could try non-daemon mode and examine the exit code. In my
case, I get this upon authentication failure.
Thanks for the pointer. I did some Google search in the archive before I
posted myself, but could not find anything relevant.
I added "set syslog" to my .fetchmailrc file, and fetchmail-5.9.11 daemon
no longer dies.
When I invoke fetchmail-5.9.11 daemon with double -v option, it also
works, but the console window is cluttered up with an error message in
effect to "procmail[]: Descriptor 1 was not open" on every poll.
Yeah, this is of course not the way you want to run fetchmail - I did
this just for testing.
I can't explain why adding "set syslog" fixes the daemon's behavior.
Since my last posting, I've found that the fetchmail daemon *always*
dies (silently) if I force it to re-read `.fetchmailrc', be it due to
a trivial modification or even just a `touch'.
I was falsely under the impression that authentication failure was the
reason for the crash; in order to test auth fail behaviour, I
intentionally hosed the password in `fetchmailrc', but it was actually
the re-reading bug that killed it. If I instead change the remote
password, the daemon keeps going.
The reason for silent death in case `.fetchmailrc' is *broken* is
explained in the man pages ("Bugs and known problems" near the end),
but the fact that it dies although `.fetchmailrc' is correct must be
considered a bug.
I have now made a self rescheduling at-job that monitors fetchmail - a
real clumsy solution, but it works. (Hmm, I might just as well use the
`at' mechanism *instead* of daemon mode, checking the exit code each
time.)
BTW, the release string of the "fetchmail -V "printout still reads 5.9.0.
This can be confusing in my opinion.
Odd, mine says `5.9.11' as expected (or actually `5.9.11+NTLM' in my
case).
Cheers
--
Björn Wirén