From: Nick Gilbert [mailto:[EMAIL PROTECTED] > > For some reason, spamassassin is behaving differently when it scans > real incoming mail compared to when I run it manually with the -t > option from the console. > > When a real message is scanned, it doesn't seem to apply any > bayesian scoring to the message (even though it's all trained up > with thousands of messages as the user 'nick' on my box). > > When I run it manually, it correctly identifies the message as spam > as it gets an extra 3.5 points by having a 99% probability of being > a spam. > > I'm running SpamAssassin via SimScan (which I presume is probably > the cause of the problem). > > Any ideas how I can fix this or is this a problem for the SimScan > list?
I don't know anything about SimScan, but the problem you describe is common enough. You are training bayes and running your manual tests as the user 'nick'. However, when a real message is scanned, it is probably running as a different user and using a different (untrained) bayes database. What user does SimScan use to run SA? Bowie
