On Sunday 28 Mar 2004 13:41, Rudi van Houten wrote:
> Hi all,
>
> I am in the process of moving from postfix with 3000+ filterrules to
> SpamAssassin under amavisd-new. I already run amavis for virus
> scanning. My test configuration runs nicely, only the Bayes filtering
> doesn't start:
>
> debug: ignore: test message to precompile patterns and load modules
> debug: using "/local/perl5/share/spamassassin" for default rules dir
> debug: using "/local/perl5/etc/mail/spamassassin" for site rules dir
> debug: using "/local/amavisd-new/.spamassassin/user_prefs" for user prefs
> file debug: bayes: no dbs present, cannot scan:
> /local/amavisd-new/.spamassassin/bayes_toks
> debug: Score set 1 chosen.
> debug: Initialising learner
> debug: bayes: no dbs present, cannot scan:
> /local/amavisd-new/.spamassassin/bayes_toks
>
> I expect that I must prime the Bayes database with spam and non-spam
> messages as I did with bogofilter. But I couldn't find how to do
> this. I presume I have to compile amavisd/amavis/spamd etc to get
> sa-learn, that was not mentioned in the amavisd-new INSTALL documents
> and the CPAN spamassasin installation. But even then it is not clear
> to me what I have to do. Can someone point me to documentation I
> missed?

As documented in the amavis .spamassassin/user_prefs file bayes doesn't work 
until it's been fed 200 spam mails and 200 ham mails. You need to feed them 
via sa-learn called by the user amavis runs under, in my case 'amavis'. Can't 
remember where I read the documentation about that, maybe in the list 
archives, which are a useful resource.

-- 
Regards,

Bob

Reply via email to