[Mailman-Developers] 'empty module name' error and shunting

Nadim Shaikli shaikli at yahoo.com
Mon Aug 18 23:38:41 EDT 2003


--- Larry Price wrote:
> On Wednesday, August 13, 2003, at 04:56  PM, Nadim Shaikli wrote:
> > If I try to 'unshunt' the message - nothing happens and it simply pops
> > back into the qfiles/shunt dir.  I have hoards of people waiting for 
> > this to get resolved and I've tried asking anyone who would listen on
> > IRC (irc.freenode.net #mailman and/or #debian) to no avail.  The website
> > was no help at all (there is no mention of 'shunting' even and alot of
> > the info is rather dated).
> 
> A quick work around to allow mail to the list to keep flowing is to 
> move the digest mbox out of the way, and then run unshunt, this will
> allow messages that were held because of the failure of the digest
> process to go through.

So it sounds like there is potential problem in which a message does stop
others from proceeding (ie. there is blocking in some instances ?)

Then Barry Warsaw wrote:
> On Wed, 2003-08-13 at 19:56, Nadim Shaikli wrote:
> 
> > Well, in my case I'm on a debian system (code via apt-get) running
> > mailman-2.1.2-6.  I have about 12 mailing-lists all seem to function
> > without an issue except for one.  Anytime anyone mails to it, the
> > message gets "Shunned" and I see this error in the log/error file
> > (mind you the message is obsoletely fine - its as though there is
> > lock someone that is forcing all legit messages to be shunned).
> 
> One message getting shunted shouldn't affect any global state that will
> cause all subsequent messages to get shunted.  Think of them as an event
> unto themselves.
> 
> Please send me the .pck/.db pair of a shunted message.  If it's not a
> problem of your configuration or installation, I should be able to
> reproduce the bug.

Barry, as was noted in a private email to ya - the problem was solved with
the following patch (courtesy of Mohammed Elzubeir),

--- Scrubber.py.orig 2003-08-13 23:19:19.000000000 -0700
+++ Scrubber.py 2003-08-14 00:23:47.000000000 -0700
@@ -305,6 +305,8 @@
                     t = unicode(t, 'ascii', 'replace').encode('ascii')
                 try:
                     # Should use HTML-Escape, or try generalizing to UTF-8
+                   if len(charset) == 0:
+                       charset = 'us-ascii'
                     t = t.encode(charset, 'replace')
                 except (UnicodeError, LookupError):
                     t = t.encode(lcset, 'replace')

once we applied the patch and 'unshunt'ed everything went through.  We
did have a couple of messages with no Mime and Content-Type headers
(which are legal to have) and those seemed to have caused the problem.
A null charset within mailman caused our problems it seems and that
potential corner-case needs to be designed for.  As for a sample - I
tried mailing the affected list a simple "test" from this yahoo
account and it got shunted (as noted there is nothing peculiar about
the messages - mailman simply got into a state/setting where it started
shunting everything -- Larry Price notes a possible means to get into
the 'state' above).  And again, I really don't think it was/is a
configuration/installation issue since the other 12 lists work without
any complaints and/or problems.

Barry, just for grins - I will mail you a message (in private) that was
indeed shunted in case that makes a difference.

Regards and I hope the patch is of use.

 - Nadim


__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com



More information about the Mailman-Developers mailing list