[QUOTE]1) why are you passing --siteconfigpath and --configpath to spamd? (This is generally NOT needed, and usually ill advised unless you have a good reason to do so). If you add these to your spamassassin call, does it change the results?
2) when you run your spammassassin tests, have you su'ed yourself to the "spamfilt" user?
[/QUOTE] (3)
[QUOTE] We had the same problem. After lot of checks it resulted
to be the more
silly thing: a permission problem with the bayes files (the amavis user couldn't read them). Check that your amavis user can use these files.[/QUOTE]
(4)
[QUOTE]I had the same issue. I am willing to bet that it is the bayes that is not being used by spamc. If so, this is due to the location of the bayes_path in your config file. Make sure that you specify a place for that. You may need to copy /root/.spamassassin/bayes_* to another location.
For example, if you were to stick those files in /etc/mail/spamassassin/bayes/, then in your config file, put this....
bayes_path /etc/mail/spamassassin/bayes/bayes
Yes, the bayes on the end is correct. It seems like it should be bayes_path /etc/mail/spamassassin/bayes/bayes* or something like that, but it isn't.[/QUOTE]
Thanks for your hints! (1) I took this from O`reilly´s SA-Book. I will try both scenarios later: spamd without and spamassassin with paramters. (2) No, I didn´t. I try too. (I tried to optimize the results for spamc rather than lowering the results of spamassassin - but for debugging reasons we should try it) Funny enough, that the bayes database sits under /home/spamfilt/.spamassassin (-> I tried to "-u" to spamfilt for the spamc call, which had no / minor effects. Strange too that the Bayes DB is under the home of spamfilt, because I as root ran my sa-learn with "-u tschloss" (tschloss is the system user receiving the mails). Under /home/tschloss there is no Bayes-DB. (3) and (4) Investigate on this too. I have to continue later (this evening in Germany). Thx Thomas
