On Wed, Feb 10, 1999 at 11:35:49AM -0500, D. J. Birchall wrote:
This is something that baffled me for a while until I did some diagnostics.
When exim is called with the -t switch, it will read recipients from the
To:, Cc: and Bcc: headers of the message. No recipients should be specified
on the command line. If in fact any are supplied, they will be DELETED from
the recipient list.
So if you happen to call exim -t with a command line address which is also
in the message header, you will end up with a null recipient list and no
delivery will occur.
In particular, the mail user agent mutt can be a problem. It takes all
recipients from headers and passes them to sendmail (or exim) on the
command line. This issue was debated on the list for quite a while, with
mutt claiming exim was broken, and Philip saying exim was correct. I tend
to agree with Philip. In the end, mutt people provided a "patch" to use
with "broken MTAs" like exim.
The moral of the story: check how you're calling exim from your CGIs and
scripts and use exim with the -d9 switch to make it print diagnostics so
you know where you are going wrong.
> I frequently write C programs and PERL scripts for system
> automation and CGI purposes, and tend to prefer piping the
> output to "sendmail -t" rather than /bin/mail when possible.
>
> Occasionally, I've noticed Exim displaying a "roach motel"
> behavior toward the output of these scripts. Messages
> check in, but they don't check out. There's an "<=" line
> in the log showing that it was received, but for whatever
> reason, no deliver was considered necessary. :)
>
> I'm just curious as to what the top 2 or 3 stupid mistakes
> that cause this would be, in language-independent terms. :)
--
Anand
System Administrator
Africa Online Ltd
http://www.anand.org
--
*** Exim information can be found at
http://www.exim.org/ ***
