On 12 Jun 2018, at 3:34, Reio Remma wrote:
Hello!
I just noticed *autolearn=ham* for a message with a positive spam
score. Is that normal?
No, but it is also not especially remarkable. The final operative score
is not the score that is used to determine autolearning.
bayes_auto_learn_threshold_nonspam is set to a negative figure in the
SpamAssassin config file.
The fact that it is "a negative figure" is pretty much useless
information. That it were -0.001 or -20 would be useful information.
X-Spam-Flag: NO
X-Spam-Score: 2.32
X-Spam-Level: **
X-Spam-Status: No, score=2.32 tagged_above=-3 required=3 tests=[
BAYES_50=0.8, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1,
DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, MAILING_LIST_MULTI=-1,
NO_DNS_FOR_FROM=1, TXREP=1.719] autolearn=ham autolearn_force=no
Your configuration is rather unusual (3 is a VERY low threshold) but
that's not why this got autolearned as ham. The score used for autolearn
thresholds does not use Bayes or TxRep/AWL scores. Detailscan be found
with 'perldoc Mail::SpamAssassin::Plugin::AutoLearnThreshold'
