Matt Kettler wrote:
Mitchell Hudson wrote:
Hello there,
I have spamassassin loaded and is running pretty well, it's supposed
to be using bayes and I can't find any errors that would tell me why
it's not, but it's not. When I do a debug log there are no db
connection errors, in fact it's auto-learning just fine. I've put in
a few thousand learning e-mails of both spam and ham so I know it's
over the min limit. Basically everything looks like it's supposed to
except there's no bayes header in the spamassassin headers and when I
tail -f /var/log/spamd/current when it says which tests it runs bayes
is never present. So basically I'm hoping there's a secret ninja
option that says 'no really I would like to use bayes'. In any case
thoughts would be appreciated.
spamassassin -V:
SpamAssassin version 3.2.0
I do use a control panel called hsphere, but it calls spamassassin
like this:
/usr/bin/perl -T -w /hsphere/shared/bin/spamd --max-children=2
--max-conn-per-child=5 --nouser-config --sql-config
--username=vpopmail --socketpath=/var/hsphere/mail/spamd
--socketowner=vpopmail --socketgroup=vchkpw --socketmode=770
--syslog-socket=none
manually calling spamassassin doesn't change the no bayes issue.
have you tried:
sa-learn --dump magic
Just to confirm there are as many messages as you think?
Hit send too soon..
Another thing to try is bayes debugging.
spamassassin -D bayes <message.txt
That should un-burry all the bayes messages from all the other debug,
and generates more bayes debug than the default -D does.
