So, I see three options to fix this, which one to use isn't my call to decide:
1. patch spamassassin package, to make sure sa-learn doesn't complain if use_bayes is set to 0
2. patch amavisd-new package, to not run sa-learn if use_bayes is set to 0. There should be a way to query SpamAssassin's configuration from a script - if there is, just use it.
3. add /etc/default/* parameter to disable spamassassin's Bayes filtering, which would require subsequent changes to amavisd-new (don't run sa-learn - but beware of remaining cleanups when this option is set to 0 after some use) and spamassassin (to translate the /etc/default/* to use_bayes 0).
I personally prefer #2 as the cleanest solution: amavisd-new owns the offending cronjob, so amavisd-new is the package that also has to deal with undesired side effects.
So, I see three options to fix this, which one to use isn't my call to decide:
1. patch spamassassin package, to make sure sa-learn doesn't complain if use_bayes is set to 0
2. patch amavisd-new package, to not run sa-learn if use_bayes is set to 0. There should be a way to query SpamAssassin's configuration from a script - if there is, just use it.
3. add /etc/default/* parameter to disable spamassassin's Bayes filtering, which would require subsequent changes to amavisd-new (don't run sa-learn - but beware of remaining cleanups when this option is set to 0 after some use) and spamassassin (to translate the /etc/default/* to use_bayes 0).
I personally prefer #2 as the cleanest solution: amavisd-new owns the offending cronjob, so amavisd-new is the package that also has to deal with undesired side effects.