Hi all, I've run into some messages recently that consistently send spamassassin into 46+ second cycles trying to classify them. Looking at the debug messages from spamd things seem to grind to a halt just as it's going into it's Bayesian stuff.
I'm running a good number of the tests from rules_du_jour, but even still, messages usually only take a couple of seconds for analysis. There's nothing obvious about what's causing these particular messages to be so slow - can I forward an example to the list to see how well your setup does against it? 2097 bytes, 55 lines. More generally, assuming that this _is_ a problem with Bayes, is there maintenance outside of the automatic maintenance performed by spamassassin by default that I can be doing to improve performance? Also, is there some way to get more verbose debugging information, like a test-by-test breakdown? Thank you, -- Mick Szucs <[EMAIL PROTECTED]> Onramp Network Services Inc.
