spamassassin --report does not train BAYES

Marc mailscanner_list at hohestieg.net
Thu Aug 11 14:30:52 UTC 2016


Dear list,

I have recently discovered some strange behavior.

If I use 'spamassassin -r < {spammessage}' reporting to DCC/Razor/Pyzor
*seems* to work according to debug. Bayes learning *should* also work
(as I made sure that 'bayes_learn_during_report' is set to 1, which is
the default anyway).
However, checking with 'spamassassin -t < {spammessage}' only gives
BAYES_50.

If I afterwards perform 'sa-learn --dbpath $DB_DIR --forget
{spammessage}' and 'sa-learn --dbpath $DB_DIR --spam {spammessage}',
things work out and I get a BAYES score of 1.0000.

======
Setup:
    Mailscanner version 4.84.5
    SpamAssassin version 3.4.1
    Perl version 5.18.4
    ----
    DB_DIR=/volume1/MailPlus/spamassassin
    local.cf: bayes_path              /volume1/MailPlus/spamassassin/bayes
   
=====

Any clues?

Thanks,
  Marc

-- 
bait: <a href="mailto:mailbait at hohestieg.net">mailbait at hohestieg.net</a>
also bait: <a href="http://www.hohestieg.net/notforhumans.html">http://www.hohestieg.net/notforhumans.html</a>



More information about the MailScanner mailing list