Am 17.06.2015 um 17:44 schrieb Dieter Scholz:
I have a new mail setup using Debian Jessie with postfix, amavisd-new, amavisd-milter and spamassassin. I use the versions from the Jessie repository. My setup does pre-queue filtering, so Spamassassin does the checks while the mail is not finally queued. Postfix calls Amavis through the milter interface.My problem is: The bayes filter does (auto-)learn ham mails but no spam mails. In my logs I found spam mails that have a very high score and should be autolearned. I think my bayes setup is correct, because ham mails are learned as expected. What could be the problem? Are spam mails not learned because of the pre-queue setup? What information do you need for helping me? Any ideas?An additional note: The ham case looks like: spam-tag, <XXX> -> <YYY>,<ZZZ>, No, score=-0.578 tagged_above=-99 required=5 tests=[HTML_MESSAGE=0.001, RP_MATCHES_RCVD=-0.58, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
you need to find out how these values are set * bayes_auto_learn_threshold_nonspam -0.001 * bayes_auto_learn_threshold_spam 9.0
signature.asc
Description: OpenPGP digital signature
