Re: Sending mail succeeded but signaled failure

[Justus Winter <justus@sequoia-pgp.org>]

David Bremner <david@tethera.net> writes:

> Justus Winter <justus@sequoia-pgp.org> writes:
>
>> Hello,
>>
>> I just embarrassed myself a little by sending the same mail over and
>> over again.  The reason for that is that notmuch-emacs signaled failure,
>> i.e. it displayed an error message in the status buffer and didn't close
>> the compose buffer, yet it did in fact send the mail.
>>
>> I suspect that my configuration has to do with that and someone is
>> trying to be helpful.  So I use msmtp with the authentication password
>> encrypted using OpenPGP.  Then, I use 'gpg --no-tty -q -d ...' as
>> msmtp's passwordeval function.  Now, my OpenPGP key has expired, but
>> that doesn't stop GnuPG from decrypting the secret, and in fact it
>> returns the status code 0.  It also prints
>>
>>   gpg: Note: secret key 08CC70F8D8CC765A expired at Mon 25 Jul 2022 05:31:26 PM CEST
>>
>> to stderr, which is picked up by notmuch-emacs, it says
>>
>>   sending...failed to gpg: Note: secret key 08CC70F8D8CC765A expired at Mon 25 Jul 2022 05:31:26 PM CEST
>>
>> in the status buffer while the compose buffer stays open.
>>
>> I suspect that this is not notmuch's fault, but I don't know where else
>> to turn to with this bug report.
>
> I guess the first step is to see if you can duplicate the problem with
> plain M-x message-mode. The mechanics of sending should be identical in
> notmuch-message-mode unless (surprise!) I misremember something.

It does indeed happen with the plain message mode.  And I think I have
identified the code in emacs that turns any output, stdout and stderr,
into errors:

% cat -n emacs/lisp/mail/sendmail.el
[...]
  1343			     (exit-value (apply #'call-process-region args)))
  1344			(cond ((or (null exit-value) (eq 0 exit-value)))
  1345			      ((numberp exit-value)
  1346	                       (setq error t)
  1347			       (error "Sending...failed with exit value %d" exit-value))
  1348			      ((stringp exit-value)
  1349	                       (setq error t)
  1350			       (error "Sending...terminated by signal: %s" exit-value))
  1351			      (t
  1352	                       (setq error t)
  1353			       (error "SENDMAIL-SEND-IT -- fall through: %S" exit-value))))
  1354		    (or fcc-was-found
  1355			(error "No recipients")))
  1356		  (if mail-interactive
  1357		      (with-current-buffer errbuf
  1358			(goto-char (point-min))
  1359			(while (re-search-forward "\n\n* *" nil t)
  1360			  (replace-match "; "))
  1361			(unless (zerop (buffer-size))
  1362	                  (setq error t)
  1363	                  (error "Sending...failed to %s"
  1364	                         (buffer-substring (point-min) (point-max)))))))

Apparently, that behavior goes back to the initial checkin of that
file.  I refuse to believe that Dan and me are the only ones having
problems with that in 30 years... I'll report it upstream.

Best,
Justus