[Spambayes] Messages avoid training with the proxy version

[Tony Meyer]

> I'm using SB at home with Thunderbird (TB), both latest versions, and
> I've recently noticed something really weird.
>
> 1. I start TB, click on 'Get mail' button
> 2. TB says '7 messages', then it goes on and downloads 7 messages.
> 3. Next I go to the 'review messages' page of SB proxy, and there I  
> see
> only 3 untrained messages (some unsure, some ham, some spam). I do the
> training.

Have a look at one of the messages that doesn't end up on the review  
page, and see if it has an "X-Spambayes-Exception" header.  It's  
possible that they are failing to parse (or something else is  
periodically wrong).

A few days back, another user reported getting quite a few spam where  
SpamBayes would fail on:

"""
X-Spambayes-Exception: Traceback (most recent call last):
   .  File "sb_server.pyc", line 498, in onRetr
   .  File "spambayes\message.pyc", line 344, in setPayload
   .  File "email\__init__.pyc", line 52, in message_from_string
   .  File "email\Parser.pyc", line 75, in parsestr
   .  File "email\Parser.pyc", line 64, in parse
   .  File "email\Parser.pyc", line 240, in _parsebody
   .  File "email\Parser.pyc", line 75, in parsestr
   .  File "email\Parser.pyc", line 64, in parse
   .  File "email\Parser.pyc", line 206, in _parsebody
   .BoundaryError: No terminating boundary and no trailing empty line
"""

The solution to this is for me to try harder to find time to get  
1.1a2 out.  1.1 uses Python 2.4, which means version 3.0 of the email  
package.  It handles any malformed message in such a way that we'll  
be able to keep going.  Before that, we had to add all of the special  
cases like this one to the SpamBayes code itself.  I could do that  
for this one, but getting that into a 1.0.5 seems less likely than  
getting 1.1a2 out.

If it becomes a big problem for anyone, they can always run the 1.0.4  
(or CVS) source with Python 2.4, and that would fix it.

=Tony.Meyer

-- 
Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes.
http://www.massey.ac.nz/~tameyer/writing/reply_all.html explains this.