I recently installed spamassassin 2.63, but I cannot make it use the bayesian auto-learner (I never see: "autolearn=yes" in mail headers), and I think it does not use the bayesian rules, either (I never see a test: "BAYES_*" in mail headers)

Here is my setup to do site-wide spam filtering:

Spamassassin is started from /etc/procmailrc:

| /usr/local/bin/spamassassin -P -C /usr/local/rules -x -p /etc/mail/spamassassin/user_prefs

Contents of file /etc/mail/spamassassin/local.cf:
report_safe 0
report_header 1
use_terse_report 1
defang_mime 0
rewrite_subject 1
use_bayes 1
auto_learn 1
bayes_path /etc/mail/spamassassin/bayes

(the file /etc/mail/spamassassin/user_prefs begins with the same lines)

I fed several hundred ham and spam to sa-learn; here is the sa-learn --dump magic:
0.000 0 2 0 non-token data: bayes db version
0.000 0 709 0 non-token data: nspam
0.000 0 761 0 non-token data: nham
0.000 0 123473 0 non-token data: ntokens
0.000 0 1073386195 0 non-token data: oldest atime
0.000 0 1078493785 0 non-token data: newest atime
0.000 0 1078493785 0 non-token data: last journal sync atime
0.000 0 0 0 non-token data: last expiry atime
0.000 0 0 0 non-token data: last expire atime delta
0.000 0 0 0 non-token data: last expire reduction count

How to setup spamassassin to actually use the bayesian rules?

P.S. Spamassassin does a great work without the bayesian rules, too, but I payed a lot of work to sort and feed mails to sa-learn, and now I want some success in making the bayesian filter to work.

Please disregard my first post, as I found that the new spamassassin version put its config files into /usr/local/share/spamassassin instead of /usr/local/rules.

So I changed the spamassassin invocation in /etc/procmailrc to:
| /usr/local/bin/spamassassin -P -C /usr/local/share/spamassassin -x -p /etc/mail/spamassassin/user_prefs
and bayesian tests began to work.