Duplicate messages from mailman (PR#88)
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
GNU Mailman |
Fix Released
|
Medium
|
Unassigned |
Bug Description
Jitterbug-Id: 88
Submitted-By: <email address hidden>
Date: Wed, 21 Jul 1999 06:31:22 -0400 (EDT)
Version: 1.0rc2
OS: Linux - Redhat 6.0
exim feeds one message into the mailman wrapper, mailman feeds out two.
Very
occasional, only noticable on a fairly high-volume list. Example from the
"post" log:
Jul 20 18:04:48 1999 post to sluts from <email address hidden> size=542
Jul 20 18:05:03 1999 post to sluts from <email address hidden> size=542
Here's another (ignore the middle entry, just interesting that it's
processing
another between the duplicates):
Jul 20 17:04:59 1999 post to sluts from <email address hidden> size=979
Jul 20 17:10:20 1999 post to sluts from <email address hidden> size=655
Jul 20 17:13:11 1999 post to sluts from <email address hidden> size=979
Headers on received mails confirm that only one is hitting the wrapper, but
there
are two coming out, a little time apart, with different exim msg-IDs.
More details on request, am prefectly willing to run it in debug mode (if
there
is
one) if that might help.
=======
Audit trail:
Wed Jul 21 12:57:10 1999 hmeland moved from incoming to open
[http://
From: Harald Meland <email address hidden>
Subject: Re: [Mailman-cabal] Duplicate messages from mailman (PR#88)
Date: 27 Jul 1999 16:35:48 +0200
[<email address hidden>]
> After an annoyingly-timed pause, finally another duplicate. The lspace. org [195.200.1.58] U=mail P=esmtp S=1174 $abc6cda0$ 040109c0@ gastuk_ nt4.gastltd_ pd T="Re: mailman/ mail/wrapper post
mailman
> log entries:
>
> [post]
> Jul 27 11:04:56 1999 post to sluts from address@concealed size=146
pid=28264
> Jul 27 11:07:40 1999 post to sluts from address@concealed size=146
pid=28464
>
> [debug]
> Jul 27 11:04:56 1999 List 'sluts': Running 'post' under pid 28264
> Jul 27 11:07:40 1999 List 'sluts': Running 'post' under pid 28464
>
> And from exim:
>
> [root@ibm1 exim]# grep 11946S-0007Lp-00 main.log # The ID from both
posts
> 1999-07-27 11:04:56 11946S-0007Lp-00 <= <email address hidden>
> H=library.
> id=014f01bed818
[sluts] Re:
> Saturday" for <email address hidden>
> 1999-07-27 11:07:52 11946S-0007Lp-00 =>
|/home/
> sluts <email address hidden> D=mailman_alias T=address_pipe
> 1999-07-27 11:07:52 11946S-0007Lp-00 Completed
So, to sum this up:
1. At 11:04:56, Exim logs reception of the message.
2. At the same time, Mailman logs startup of the "post" script
in the
debug log.
3. Also at the same time, Mailman logs a delivery to the list from
pretty far down in the MailList.Post() method.
4. Then, at 11:07:40, Mailman logs startup of the "post" script
once
more, now with a different pid.
5. Also at 11:07:40, Mailman logs a new delivery to the list, with
the same pid as the latter "post" script.
6. Exim logs that the delivery is completed at 11:07:52. Delivery
completion means that the Mailman pipe has exited with status zero
(indicating a successful delivery).
> So, two seperate (but similar - _surely_ coincidence?)
As the printed PIDs are taken directly from calls to os.getpid(), I
think it is rather safe to assume that the similarity is coincidental.
> PIDs, which is clearly what you were looking for.
Indeed.
> I presume that this has to be exim barfing and retrying for some
> bizzarre reason?
If not exactly "barfing", then at least spawning more than one
pipe
delivery while apparently only logging one of them.
> Although I'd have expected it to reflect that in it's own
> log... perhaps the wrapper doing much the same?
I've grepped the entire Mailman source for anything that might re-exec
the "post" script, but found nothing. And, a simple re-exec
would not
be enough -- the new "post" instance would also need to be fed
the
message on it's stdin.
I haven't had the time to keep up with the Exim-users list recently,
but I (now) strongly suspect that this problem belongs there rather
than with Mailman.
If the problem is of such a nature that it might also bite other
Mailman/Exim users, please keep us (<email address hidden>)
posted on any progress/solution.
Cheers,
--
Harald