[Mailman-Developers] Current round of checkins are finished

Thomas Wouters thomas@xs4all.net
Wed, 10 May 2000 22:32:42 +0200

On Tue, May 09, 2000 at 04:39:24PM -0400, bwarsaw@python.org wrote:

> I've checked in all my current modifications to make losing messages
> much less likely.  I'm sure there are bugs lurking about, so be
> careful if you're running the CVS snapshot.  I'm going to spend a few
> hours over the next couple of days testing things so watch for more
> patches.  If you do play with the snapshot and find problems, please
> let me know.  Figure if I get some time, there will be a beta this
> weekend.

I was very pleasantly surprised by your patches :-) I had something similar
but much less ambitious lurking in my devel tree, which used a
QueuemessageError (which contained a list of failed recipients) to notify
the pipeline mechanism that delivery failed and should be queued. I wasn't
very happy with it, so I'm glad I can chunk it now :)

I noticed one lurking bug though: the code uses 'str(msg)' to dump a message
to a file. This drops the unixfrom line, causing pipermail/hyperarch to barf
on the message (in a silent way, unfortunately.) I'm not sure if changing
Mailan.Message.Message.__str__() to add the unixfrom line would be a good
idea, though, for code-consistency reasons. Even though most rfc822 stuff
handles the unixfrom line fine, it might create some strange behaviour when
passing a regular rfc822 message, for instance.

(Seeing how Mailman.Message reads in the entire body, I'd think it entirely
possible that some parts of Mailman, say the Archiver, will handle regular
rfc822 messages. Then again, it could be a seperate subclass of
rfc822.Message that only overrides __str__, so it's not tha big an issue.)

I'll test the new code when I finish merging in my own changes, and deleting
the obsolete ones ;-P

Thomas Wouters <thomas@xs4all.net>

Hi! I'm a .signature virus! copy me into your .signature file to help me spread!