A bug in 2.6 caused messages which hit BAYES_99 to be learned as ham, this has been fixed, you should upgrade.
For reference, there was no bug per se. The fact that the message hit BAYES_99 did not cause it to be learned as ham.
However, newer versions of SA, as an enhancement, will prevent autolearning if the autolearning will go very strongly against the previous bayes training. (ie: it will skip autolearning a BAYES_99 message as ham, or a BAYES_00 message as spam).
It's kind of going with the principle of "if you're getting mixed signals, don't learn", which makes sense
You can read all about it here:
http://bugzilla.spamassassin.org/show_bug.cgi?id=2437
------------------------------------------------------- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk