>> Looks like your training database is corrupt. Just zap it and
>> retrain from your mailboxes of saved spam and ham.
Mark> Ok, my database is corrupt again overnight. All I do is train it
Mark> overnight once. Why would the training corrupt it?
Most of the database types can only have one process accessing them at
once. If you happen to have two writers your database is almost guaranteed
to be corrupt. If you have a reader and a writer it's still possible that
the reader will see the database in a transitional state which is
inconsistent with the expectations of the reader.
What type of database storage are you using? I seem to recall you're using
sb_filter.py. Make sure mail isn't processed when you're retraining. In my
procmailrc file I use sb_bnfilter.py:
SCORE="sb_bnfilter.py -a 30 -A 250"
...
:0 fw:scorelock~
| $SCORE
When retraining I make sure that incoming mail isn't processed. If
necessary, I kill the long-running sb_bnfilter.py process first.
Skip
_______________________________________________
SpamBayes@python.org
http://mail.python.org/mailman/listinfo/spambayes
Info/Unsubscribe: http://mail.python.org/mailman/listinfo/spambayes
Check the FAQ before asking: http://spambayes.sf.net/faq.html
