[Mailman-Users] Mailman/pipermail/MHonArc integration issue

[Richard Barrett]

On 1 May 2004, at 19:40, Wiley Wimberly wrote:

>
> Hi, I am using the Mailman/pipermail/MHonArc integration patchs found 
> at http://www.openinfo.co.uk/mailman/index.html and have been seeing 
> the following issue. When I run bin/arch --wipe listname it correctly 
> builds the html for each month section unless there is only a single 
> message for the month. The output shows:
>
>
> figuring article archives
> 2004-May
>
> Warning: Could not parse date for message
>          Message-Id: 
> <f7589c774a9bf75f62148a712f1140fd at NO-ID-FOUND.mhonarc.org>
>
> This only seems to happen if there is just one message for that month.
>
> I also see the same error in the error log on all incoming messages 
> that are sent to the archiver.
>
> Warning: Could not parse date for message
>          Message-Id: 
> <4173763d54406fe2a588db538ad84586 at NO-ID-FOUND.mhonarc.org>
>
> The message shows up in the html, but the subject and sender are 
> missing from the toc.
>
> 	• 	[no subject] ,  Unknown - Sat May 01 14:28:54 2004
>
> Also all of the header lines are missing from the message page. I just 
> get a heading of  [no subject].
>
> I am using mailman 2.1.4 with the following patches which all seemed 
> to apply cleanly.
>
> modinc-2.1.4-0.1.patch
> indexing-2.1.4-0.1.patch
> htdig-2.1.4-0.1.patch
> mhonarc-2.1.4-0.1.patch
>
>
> Is anyone else seeing this issue? I have been beating on this for a 
> few days and any suggestions will be greatly appreciated. Thanks.

I have not seen this problem during development and testing of the 
mhonarc integration patch you are using. Nobody has raised a similar 
problem with me but I have no idea how many users (if any, besides 
yourself) there are of the patch. I am running it on servers I maintain 
which have been getting fairly heavy use without seeing this problem.

The MHonArc FAQ provides an explanation for why the message is 
generated but I cannot immediately associate this with an obvious 
defect in the mhonarc integration patch. See MHonArc FAQ at:

     http://www.mhonarc.org/MHonArc/doc/faq/usage.html#parsedate

What version of MHonArc are you running? The patch was developed and 
tested using MHonArc 2.6.8

The say you get this problem when you only have one message in a month 
(I assume this is the archive period for the list?) when running 
bin/arch or when new messages are being handled. This is significant in 
that under these circumstances a single message is piped to the spawned 
process running the mhonarc script, by the patched Mailman/pipermail 
code, rather than the spawned process being handed multiple messages 
written to a temporary mbox file by the patched Mailman/pipermail code. 
Why this should make a difference is not immediately obvious to me.

With MHonArc 2.6.8 source, the actual error message appears to be 
generated around line 818 in the file 
lib/perl5/site_perl/5.6.1/mhamain.pl (under whatever directory path you 
have installed MHonArc in). From what you say about the HTML generated 
for these messages, the failure mode appears to be a general failure 
for MHonArc to identify any headers in the message; in the info you 
cite the message id, date, subject and other headers appear not to have 
been found by the mail header parser used by MHonArc.

On your system, what OS and version, Python version and how installed, 
how was Mailman installed, MHonArc version and how installed?

I'll take a look at the MHonArc 2.6.8 code in abit more detail but I am 
limited on the time and effort I can spend on this; I have some urgent 
paid-work commitments I must meet right now.

I can provide you with some diagnostic code to try and confirm what the 
patched Mailman is piping to MHonArc in order to identity which side of 
the Mailman/MHonArc line the search for a problem solution needs to 
concentrate on, if you want.