I am using SA 3.0.1. All seems ok, but auto-learn is unlearning emails that are marked as spam in the Baysian filter.
Running the following registers the email in the Baysian filter ok: spamassassin -D -r sample-spam.txt But when I scan the message: spamassassin -D -t < sample-spam.txt > spam.out I get the following output: <snip> debug: auto-learn: currently using scoreset 1. debug: auto-learn: message score: -2.82, computed score for autolearn: -2.82 debug: auto-learn? ham=0.1, spam=12, body-points=0, head-points=-2.82, learned- points=0 debug: auto-learn? yes, ham (-2.82 < 0.1) <snip> debug: bayes: found bayes db version 3 debug: [EMAIL PROTECTED]: already learnt as opposite, forgetting first debug: refresh: 25784 refresh /home/sabedard/.spamassassin/bayes.lock But I want to keep this message learned as spam. It seemed to me disabling auto-learn in the $HOME/.spamassassin/user_prefs file might fix this, but this setting seems to be ignored as the following still appears in the debug output: debug: auto-learn? yes, ham (-2.82 < 0.1) But the config file has the following: auto_learn 0 Any ideas on this? Is this a bug or my bad?
