Hi! Please bear with me, I'm new to this list. If I repeat FAQ I'll happily take pointers to the relevant document(s).
I'm running SA 2.55. Sadly, the server is underpowered, so I can't let sa-learn learn from > 10.000 messages there. I've populated the bayes_* - files on another machine, and, via a step converting Berkeley DB v7 to v5, copied them to the mailserver. sa-learn seems happy updating the files, however, I never get any BAYES-rules matching. Running sa-learn with -D I see debug: debug: Only 8 spam(s) in Bayes DB < 200 I've trained sa-learn with _a lot_ messages (half ham/spam) in this test, and bayes_msgcount is 1409 bytes. 8 is, I guess, what SA has auto-learned since I copied the bayes_* - files. When I run SA on the machine I created the bayes-files on, I see debug: bayes corpus size: nspam = 14385, nham = 15718 Any hints on how I could get this working? Maybe I did something wrong when converting from Berkeley DB v7 to v5 (I just dumped the db locally and imported it on the mailserver)? It's the same SA, but not the same perl-version & OS-release on the 2 machines. Any hints would be appreciated. cheers+tia, &rw -- -- "Robots are fascinating because they are like people, -- and people are disgusting narcissistic beasts." -- http://ftrain.com/ah_beautiful_robot.html
pgp00000.pgp
Description: PGP signature
