On Fri, May 13, 2005 at 08:35:08AM -0700, Steven Manross wrote: > On that same note, > > I was playing around with BAYES on a MS SQL Server... > (I followed the directions in the readme files and tweaked the mysql.sql > code to work for MS SQL in creation of the tables.)
Ahhh, most excellent, I don't know of anyone running on MS SQL Server. I'm glad to hear that you got pretty far. If you've got a .sql file that seems to work (once we debug your current problem of course) we should get it added to the source tree. > > The import worked fine (no errors that I could see). > > sa-learn --D --backup >bayes-backup.txt) It may not be wise to run with -D when backing up, it might be a little verbose to be useful. Can you cut-and-paste the first 10 or so lines of this file? That will give an indication if they look ok. > Mod the CF files > > sa-learn -D --restore bayes-backup.txt) > Might also be helpful, once you've got a decent bayes db, so goahead and do a --backup on this one and make sure everything looks sane. If you do, go ahead and cut-and-paste the first 10 or so lines as well. > But no bayes tests seem to hit since modifying the local.cf for the new > bayes store.. > > I can see the SQL traffic hitting the database, but nothing seems to be > hitting BAYES_** at all. > > Any ideas? > It's possible that you just aren't hitting enough significant tokens, but unlikely depending on your database size. One quick question, what username did you run sa-learn as? Are you sure that the entries in bayes_vars are under the username that you have specified in the bayes_sql_override_username option? Run a msg through with spamassassin -D and paste the debug output, it might give us an indication. Also, any other tips, possible changes to README.bayes, you can provide for running under MS SQL Server would be great, I'll see about getting those folded into the source tree. Michael
pgp8kIdUC8qi9.pgp
Description: PGP signature
