Manual learning is not working according to bayes db #123
Description
After a fresh install there should be an empty bayes db, by marking quarantined messages as spam or ham this db should fill slowly but steadily.
This does not happen according to sa-learn --username=vmail --dump magic
So I decided to make an initial learning with
sa-learn --ham --username=vmail /usr/share/doc/spamassassin/examples/sample-nonspam.txt.gz
and
sa-learn --spam --username=vmail /usr/share/doc/spamassassin/examples/sample-spam.txt
Resulting in:
sa-learn --username=vmail --dump magic
0.000 0 3 0 non-token data: bayes db version
0.000 0 1 0 non-token data: nspam
0.000 0 1 0 non-token data: nham
0.000 0 547 0 non-token data: ntokens
0.000 0 987802486 0 non-token data: oldest atime
0.000 0 1058995800 0 non-token data: newest atime
0.000 0 0 0 non-token data: last journal sync atime
0.000 0 0 0 non-token data: last expiry atime
0.000 0 0 0 non-token data: last expire atime delta
0.000 0 0 0 non-token data: last expire reduction count
Now I again marked mails in quarantine spam or ham and was expecting to see changes in the bayes db. After a couple of mails and some minutes waiting I just saw the same output.
Nothing changed in the bayes db. This seems not right.
I activated debug and started modoboa via manage.py runserver but there was no exception. Everything seems normal.
Let me know if I can provide more information.