Hello John, Monday, October 3, 2005, 1:11:23 PM, you wrote:
JH> Only clue I have is that if I run sa-learn, it takes minutes to respond at JH> all, and in the output appears "Parsing of undecoded UTF-8 will give JH> garbage when decoding entities at JH> [...]/Mail/SpamAssassin/HTML.pm line JH> 182" - although I suspect this message is a red herring... Agreed with the red herring. I'm thinking your Bayes database may be hosed, and sa-learn is trying to kick off a cleanup, which causes your problem. And any SA scan of an email will do the same thing. If you turn off Bayes briefly, does the problem go away? Can you afford to delete the Bayes database files and reinitialize them by doing sa-learn on 200+ confirmed spam and 200+ confirmed non-spam? That might be your best solution. Bob Menschel
