Oddity with mail

Oddity with mail

[Jack Twilley]

Recently I started suffering a bit of email trouble.

I would send messages, and they would seem to go away just fine, but
they'd never get to their destination, and they'd never bounce.

I can send mail with the command-line tool just fine, though.  How
odd.

Here's a snippet of the logs from a recent test that succeeded from
the command line:

Mar 10 00:17:40 duchess sendmail[21511]: f2A8HeH21511: from=jmt, size=53, class=0, nrcpts=1, msgid=<200103100817.f2A8HeH21511@duchess.twilley.org>, relay=jmt@localhost
Mar 10 00:17:50 duchess sendmail[21513]: f2A8HeH21511: to=jmt+usenet@tbe.net, ctladdr=jmt (1001/1001), delay=00:00:10, xdelay=00:00:10, mailer=relay, pri=30053, relay=mulan.twilley.org. [10.74.84.254], dsn=2.0.0, stat=Sent (f2A8IKf21223 Message accepted for delivery)

Here's a snippet from a failed message:

Mar 10 00:18:22 duchess sendmail[21524]: f2A8IMF21524: Authentication-Warning: duchess.twilley.org: jmt set sender to jmt@tbe.net using -f
Mar 10 00:18:22 duchess sendmail[21524]: f2A8IMF21524: from=jmt@tbe.net, size=821, class=0, nrcpts=0, msgid=<200103100818.f2A8IMF21524@duchess.twilley.org>, relay=jmt@localhost

I'm used to the warning -- that's fine -- but the interesting
difference I found was that "nrcpts" value.  In the command-line
example, it's equal to one, and in the gnus example, it's equal to
zero.

Now I haven't changed anything as far as I know other than upgrading
Gnus.  It's possible I might have a messed-up sendmail.cf, but I don't
think so.

Any suggestions or hints?

Jack.
-- 
Jack Twilley
jmt at tbe dot net
http colon slash slash www dot tbe dot net slash tilde jmt slash

Re: Oddity with mail

[Kai Großjohann]

Gnus is supposed to call `sendmail -oi -t' and the `-t' tells sendmail
to glean the recipients from the header.

Can you use `ps' to see if that's the case on your system?  And does
it work to call `sendmail -t' from the command line?

kai
-- 
Be indiscrete.  Do it continuously.

Re: Oddity with mail

[Jack Twilley]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

Kai> Gnus is supposed to call `sendmail -oi -t' and the `-t' tells
Kai> sendmail to glean the recipients from the header.

I could have trolled the code to determine that.  Duh.  Must have been
tired.

Kai> Can you use `ps' to see if that's the case on your system?  And
Kai> does it work to call `sendmail -t' from the command line?

I set up a tight loop to grep for sendmail incarnations, and it caught
this one when I sent another failed message:

/usr/sbin/sendmail -oi -f jmt@tbe.net -oem -odb -t

When I run this command by hand with the following input, it works:

--begin input--
To: jmt@thud.tbe.net
Subject: Hi.

Bye.
--end input--

How do I capture what Gnus is sending to this process?

Kai> kai -- Be indiscrete.  Do it continuously.

Jack.
-- 
Jack Twilley
jmt at tbe dot net
http colon slash slash www dot tbe dot net slash tilde jmt slash

Re: Oddity with mail

[Kai Großjohann]

On 10 Mar 2001, Jack Twilley wrote:

> How do I capture what Gnus is sending to this process?

With M-x debug-on-entry RET message-send-mail-with-sendmail RET, I
guess.  Maybe putting point inside that defun and then doing M-x
edebug-defun RET is a more user-friendly way.

You just single-step through the function until you see that it's about
to invoke sendmail.  Then you look at the buffer that's to be sent to
the process.

kai
-- 
Be indiscrete.  Do it continuously.

Problem identified - xemacs package? (was Re: Oddity with mail)

[Jack Twilley]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

Jack> How do I capture what Gnus is sending to this process?

Kai> With M-x debug-on-entry RET message-send-mail-with-sendmail RET,
Kai> I guess.  Maybe putting point inside that defun and then doing
Kai> M-x edebug-defun RET is a more user-friendly way.

The former worked just fine.

Kai> You just single-step through the function until you see that it's
Kai> about to invoke sendmail.  Then you look at the buffer that's to
Kai> be sent to the process.

And I found the problem.  For whatever reason, my incarnation of Gnus
under XEmacs is calling an old version of message.el which has a bug
in message-add-header.  I tried removing the gnus package, and that
broke everything.

Any chance a new XEmacs package could be built?  Or, failing that,
could someone who understands these packages add the necessary targets
to the Makefile so those of us who need them can do it ourselves?

Kai> kai -- Be indiscrete.  Do it continuously.

Jack.
(whee.)
-- 
Jack Twilley
jmt at tbe dot net
http colon slash slash www dot tbe dot net slash tilde jmt slash