[Savannah-hackers-public] Re: [gnu.org #360160] Fw: Cron <address@hidden> /usr/bin/python -S /var/mailman/cron/senddigests

[Sylvain Beucler]

On Wed, Apr 16, 2008 at 03:20:09PM -0400, Ward Vandewege via RT wrote:
> > [beuc - Wed Apr 16 02:52:09 2008]:
> > 
> > On Tue, Apr 15, 2008 at 05:11:04PM -0500, Karl Berry wrote:
> > >     address@hidden receives this error on a regular basis (I'd say
> > every
> > >     day).
> > >
> > >     Any clue?
> > >
> > > Basically, no.
> > >
> > > We've discussed it before, at some length.  It is obviously a bug
> > > somewhere in mailman but no one has had time/interest to track it
> > down.
> > > Joshua had the thought that maybe turning off the digest delivery
> > option
> > > would stop it, but it didn't.
> > >
> > > I've now altered /var/mailman/cron/senddigests.karl to once again
> > skip
> > > libtool-patches (along with gnu-arch-users and myexperiment-
> > topsecret,
> > > for which the problem also happened).
> > >
> > > A similar thing happens with checkdbs on bug-coreutils.
> > >
> > > If you or anyone wants to delve into it, feel free.  The original
> > > scripts in /var/mailman/cron/*.orig; my changes in *.karl are
> > entirely
> > > trivial.
> > 
> > Hi,
> > 
> > Here's the backtrace (by executing the manually w/o try/catch):
> > 
> > >>> mlist.send_digest_now()
> > Traceback (most recent call last):
> >   File "<stdin>", line 1, in ?
> >   File "/var/mailman/Mailman/Digester.py", line 60, in send_digest_now
> > > ToDigest.send_digests(self, mboxfp)
> >   File "/var/mailman/Mailman/Handlers/ToDigest.py", line 155, in
> > send_digests
> > > send_i18n_digests(mlist, mboxfp)
> >   File "/var/mailman/Mailman/Handlers/ToDigest.py", line 246, in
> > send_i18n_digests
> > > msgsubj = msg.get('subject', _('(no subject)'))
> > AttributeError: 'str' object has no attribute 'get'
> > 
> > Here's the code involved (Mailman 2.1.5):
> > 
> >    msg = mbox.next()
> > > while msg is not None:
> > >     if msg == '':
> > >         # It was an unparseable message
> > >         msg = mbox.next()
> > >     msgcount += 1
> > 
> > Here's the code in Mailman 2.1.9:
> > 
> > > while msg is not None:
> > >     if msg == '':
> > >         # It was an unparseable message
> > >         msg = mbox.next()
> > >         continue
> > >     msgcount += 1
> > 
> > (additional "continue")
> > 
> > 
> > Afaict the error occurs when 2 successive messages in
> > /var/mailman/lists/<listname>/digest.mbox are unparseable.
> > 
> > The error probably goes away next time the digest is send (in the case
> > of libtool-patches, each month only).
> > 
> > Karl, I revert your changes (senddigest and crontab -e) so we see if
> > the digest will be sent for libtool-patches (it already contains mails
> > from 16 March - 15 April, so maybe it will be sent soon).
> > 
> > 
> > Is there a reason why Mailman isn't upgraded to 2.1.9?
> 
> Lists runs a very old OS setup. It's going to be replaced at some point
> in the future. In the mean time, it's complicated to run our (heavily
> customized) Mailman setup.
> 
> I think adding that 'continue' to the codebase manually is probably the
> best way forward, if it fixes the bug.

Well that did the trick AFAICT.

Is there a place where I should note about this change?


I also added a complete traceback print in senddigests for next time.


For myexperiment-topsecret there's another error, but since nobody is
using digests there, I just removed the (2 months old) monthly digest.

Similarly the digest for gnu-arch-users is 3 (!) years old, I removed
it as well.


Apparently we need to avoid manually excluding some lists, because
this essentially postpone digest delivery indefinitely.

Cheers,

-- 
Sylvain