[spambayes-bugs] [ spambayes-Bugs-1098054 ] No more training on new mails

[SourceForge.net]

Bugs item #1098054, was opened at 2005-01-07 13:18
Message generated for change (Comment added) made by kpitt
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=498103&aid=1098054&group_id=61702

Category: pop3proxy
Group: 1.0.1
Status: Open
Resolution: None
Priority: 5
Submitted By: Alain (alain_38)
Assigned to: Nobody/Anonymous (nobody)
Summary: No more training on new mails

Initial Comment:
New messages are counted but none of them is classified
nor available for training. They are sent as ham to the
email client (Outlook Express). All these messages show
the same  following information in their headers:

X-Spambayes-Exception: Traceback (most recent call last):
 .  File "sb_server.pyc", line 475, in onRetr
 .  File "spambayes\classifier.pyc", line 190, in
chi2_spamprob
 .  File "spambayes\classifier.pyc", line 493, in _getclues
 .  File "spambayes\classifier.pyc", line 508, in
_worddistanceget
 .  File "spambayes\classifier.pyc", line 308, in
probability
 .AssertionError

This problem occurred for the first time when running 
1.0rc2. I installed 1.0.1 over it hoping that it would
fix the bug.
I have 3 PCs running pop3proxy SpamBayes (2 Windows XP,
1 Windows 2000, all Pro versions) and only one of them
has this bug. This one is running XP.

----------------------------------------------------------------------

>Comment By: Kenny Pitt (kpitt)
Date: 2005-01-07 14:36

Message:
Logged In: YES 
user_id=859086

Near as I can tell, this error occurs if a token in your
training database indicates that it has been trained more
times than the actual number of messages that you have
trained on. Since a given token is only trained once per
message, this should never happen.

What this usually comes down to is that your training data
is now out of whack and there really isn't anything that can
be done to repair it.  The only workaround, unfortunately,
is to retrain SpamBayes from scratch.  If you make regular
backups then you might be able to find a backup of your
training data before the problem occurred and go back to
that version.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=498103&aid=1098054&group_id=61702