Hello,
Attached you find the output of the sa-learn execution. The nine
spam E-Mails learned are there because I manually learned a mbox
file with spam. As you can see the ham part seems to work.
is there anyone who could help me on this or anyone who has the same
problem?
Or is it better to ask on the amavis group because SA is called from
the amavis process?
Probably, I've not heard of this happening with spamd.
Autolearning is done as a side effect of scanning, so if it's
working on ham, it should work on spam. It's hard to see how the
calling code could discriminate. What I'm wondering is if something is
removing all the high scoring spam before SA can auto-learn it.
I don't use amavis, so perhaps it's normal for there to be such a big
difference in the SA logging of rejected mail and other mail, but is
it possible that it's [mis]configured to send mail through SA twice?
I just found a configuration option in amavis $sa_dsn_cutoff_level which
was set to 10 in the Debian default config. I think this ia a possible
candidate for my problems. I changed this to 99.
I will give feedback on this.
Dieter
