I'm curious how to determine why a message gets flagged with "autolearn=no", as shown below:
Well, as far as the learner was concerned it scored only whatever NO_REAL_NAME contributes in scoreset 0 or 1. In 2.60, that's 0.339 or 0.285.
That's not high enough to learn as spam, nor low enough to learn as ham.
Poof, no autolearn performed due to midline score.
Then the AWL kicked in and deducted a large number of points.
X-Spam-Status: No, hits=-5.9 required=5.0 tests=AWL,NO_REAL_NAME autolearn=no version=2.60-spambr_20030926a
Also, what does this "version=2.60-spambr_20030926a" mean?
That's the build of SA which generated the X-Spam-Status header in question.. It's some Brazilian localized fork if I recall.
This was a veryobvious pill spam message, yet it ended up with -5.9. Any insight would be
greatly appreciated.
2.60 is now a year old has serious AWL bugs that spammers can abuse. Upgrade to something more recent.
Also, if your system isn't running 2.50-spambr, then make sure you're not doing something foolish like skipping the scan for messages with X-Spam-Status headers in them (which spammers can forge or other untrusted systems can add).
