Jesse Keating wrote:
And of course I just found the problem. email2trac's spam
detection
code is... sub par. It was seeing a 'X-Spam-Level:' header with no
content and trying to split that line. Whoops. It seems to me it
really needs to grok the X-Spam-Status line and find the score within.
Why is parsing X-Spam-Status better than counting the '*' in
X-Spam-Level though? As long as the code doesn't try to split an
empty field, I think it should be fine. (Unless there are cases where
Spamassassin inserts X-Spam-Status and no X-Spam-Level.)
I don't necessarily have time to do that today so I've hand
patched
email2trac to just catch on 'X-Spam-Flag'. If any of you would like
to work on a better patch please shoot it my way, I'll hold off on
building email2trac for a few days.
Here[1] is something only lightly tested (outside of trac, as I have
no trac install handy to test with). It might be something worth
adding along with the check you've added to catch X-Spam-Flag: Yes.
[1]
http://tmz.fedorapeople.org/tmp/email2trac-0.13-spamcheck.patch
--
Todd OpenPGP -> KeyID: 0xBEAF0CE3 | URL:
www.pobox.com/~tmz/pgp
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Total strangers need love, too; and I'm stranger than most.