On 13/02/2017 17:33, [email protected] wrote:
So I did a further dig
I replaced my dspam binary with a shell script
#!/bin/sh
echo 'tee dspam.in | dspam '$@' | tee dspam.out' > /tmp/dspam.cmd
tee /tmp/dspam.in | dspamx $@ | tee /tmp/dspam.out
Turns out the autolearn process doesn't get passed back the headers
dspam added in its result, when the autolearn is called.
Oddly the advice was to ensure SpamAssassin was the last entry on the
scam_scanners list, to ensure it got the headers added by DSPAM, so SA
does get the DSPAM headers, but the auto-learn proccess doesn't. :(
Guess I'm gonna have to learn some perl and work out wtf is going on.
While I applaud the concept of making amavisd more generic with the
spam_scanners etc. for the love of god don't do it such a way that it
breaks current features such as dspam support.
BA
