MailScanner, sa-learn and bayes_path troubles

[Matthias Kellermann]

Hello MailScanner list,

I've some trouble using sa-learn to train my SpamAssassin installation
with MailScanner.

If I use sa-learn to train the filter with ham and spam messages, it
doesn't seem to have any effect on the scoring done by MailScanner. I
also don't see where sa-learn puts its Bayes data. I thought the path
should be the same as bayes_path in
/etc/MailScanner/spam.assassin.prefs.conf but there are no files
present. It doesn't change anything if I run sa-learn with root or
postfix user (MailScanner runs as postfix user).

So, what could be the problem? Where can I tell sa-learn which path to
save it's ham/spam information? As which user I have to run sa-learn? Do
I have to add another special bayes path somewhere in the MailScanner
config files?

Thanks for your answers.

Regards,
Matthias