On Thu, Oct 16, 2025 at 10:32 AM Bill Cole <[email protected]> wrote: > > On 2025-10-16 at 09:48:23 UTC-0400 (Thu, 16 Oct 2025 09:48:23 -0400) > Ricky Boone <[email protected]> > is rumored to have said: > > > Bayes is using the Mail::SpamAssassin::BayesStore::SQL store module > > with DBI:mysql (the actual backend server is MariaDB, though). > > Oh, wait... This is BAD. See below > > $ perldoc Mail::SpamAssassin::BayesStore::SQL > NAME > Mail::SpamAssassin::BayesStore::SQL - SQL Bayesian Storage Module > Implementation > > DESCRIPTION > This module implements a SQL based bayesian storage module. It's > compatible with SQLite and possibly other standard SQL servers. > > Do not use this for MySQL/MariaDB or PgSQL, they have their own > specific > modules. > > Use the right plugin. Wipe the DB first, as you may have mangled it by > using the wrong plugin.
Agreed. I'm trying to go back and look, but I don't think there was a different option for SA 3.4 when using mysql/mariadb backends. It was running without issue for a long period of time with the same backend database (minus mariadb version changes over time), so the difference seems to have come up with SA 4.0+ that I didn't notice. That said, it "works" more than it's failing; I can't get it to fail when I debug manually. Not doubting there's an issue, though, so definitely on my list of things to try. I also have backups of the database from before pointing SA 4.0.1 to it in case I need to roll back, but wiping isn't out of the question. As for ShortCircuit, I don't have it loaded or configured, and it isn't in my manual debug attempt output (other than references to config files where the filename might match).
