Bayes problem with MailScanner

[Jan-Peter Koopmann]

Hi,

> Does it change if you do a
>          sa-learn --rebuild
> ?

I am doing this once an hour via cron job. Have a look at the debug
output: It is using the same db as the check_bayes_db AND the same db as
it is using when I am manually doing a spamassassin -t. I am 99.9999%
sure that the tokens are in the right db and are correctly learned.
That's what is so strange about this. Somehow SA interprets the db
contents differently depending on how it is called. I had a quick look
at the source and could not find anything...

Regards,
  JP