[Mimedefang] Bayesian analysis not working

[Daniel Kasak]

I'm trying to activate bayesian analysis.

My /etc/mail/spamassassin/sa-mimedefang.cf:

required_hits           7
ok_locales              en
rewrite_subject 0
report_header 1
skip_rbl_checks 0
use_bayes               1
auto_learn              1
bayes_path              /var/spool/MIMEDefang/bayes
bayes_file_mode         0666
auto_whitelist_path     /var/spool/MIMEDefang/auto_whitelist
auto_whitelist_file_mode        0666
bayes_min_ham_num       20
bayes_min_spam_num      20

Spamassassin is working, minus bayesian analysis. I get the following 2 
lines in all email:

X-Spam-Score: 0.675 () HOME_EMPLOYMENT,WORK_AT_HOME
X-Scanned-By: MIMEDefang 2.38

But nothing about bayesian analysis.

With the above settings, I have the following in /var/spool/MIMEDefang:
proto MIMEDefang # ls -l
total 1245
drwxr-xr-x  2 defang users      48 May 20 23:47 auto_whitelist
-rw-------  2 defang users   12288 May 20 23:49 auto_whitelist.dir
-rw-------  2 defang users   12288 May 20 23:49 auto_whitelist.pag
drwxr-xr-x  2 defang users      48 May 22 10:38 bayes
-rw-rw-rw-  1 defang users  626688 May 22 10:28 bayes_seen
-rw-rw-rw-  1 defang users 1236992 May 22 10:28 bayes_toks
srwxrwxrwx  1 defang users       0 Apr 13 19:25 clamd.sock
-rw-------  1 defang users       6 May 22 10:37 mimedefang-multiplexor.pid
srw-------  1 defang users       0 May 22 10:37 mimedefang-multiplexor.sock
-rw-------  1 defang users       6 May 22 10:37 mimedefang.pid
srwx------  1 defang users       0 May 22 10:37 mimedefang.sock
proto MIMEDefang # 

I created the 2 directories 'auto_whitelist' and 'bayes', but instead of using them, things seem to get dumped in the top-level directory.
Also when I run 'sa-learn' the bayes files above get created by root, so I chown then to defang. I assume there is a better way of doing this?
Maybe running sa-learn as the defang user?

Anyway, I have run 'sa-learn' and learned about 50 spam messages, and a couple of hundred ham messages.

What else should I be looking for?

Dan