SA Bugzilla – Bug 6786
near zero score for TO_EQ_FM*
Last modified: 2018-02-03 17:39:44 UTC
few days ago sa-update greatly lowered the scores for TO_EQ_FM* tests: 72_scores.cf: previous: score TO_EQ_FM_DIRECT_MX 1.650 0.659 1.650 0.659 score TO_EQ_FM_HTML_DIRECT 3.134 3.950 3.134 3.950 score TO_EQ_FM_HTML_ONLY 2.374 3.008 2.374 3.008 current: score TO_EQ_FM_DIRECT_MX 0.001 0.001 0.001 0.001 score TO_EQ_FM_DOM_HTML_IMG 0.001 0.001 0.001 0.001 score TO_EQ_FM_DOM_HTML_ONLY 0.001 0.001 0.001 0.001 score TO_EQ_FM_DOM_SPF_FAIL 0.001 2.638 0.001 2.638 score TO_EQ_FM_HTML_DIRECT 0.001 0.001 0.001 0.001 score TO_EQ_FM_HTML_ONLY 0.001 0.001 0.001 0.001 score TO_EQ_FM_SPF_FAIL 0.001 2.367 0.001 2.367 and I believe something is wrong with that since TO_EQ_FM* are nice enough to mark the spam I receive
A new sa-update hasn't been generated since 2012-02-25, how did your scores change a few days ago? Those rules are ranked pretty terribly. Best rules are ranked 1, worst are ranked 0: MSECS SPAM% HAM% S/O RANK SCORE NAME WHO/AGE 0 0.5257 0.0050 0.991 0.54 0.00 TO_EQ_FM_HTML_ONLY 0 0.3420 0.0017 0.995 0.52 0.00 TO_EQ_FM_DIRECT_MX 0 0.2501 0.0011 0.996 0.51 0.00 TO_EQ_FM_HTML_DIRECT - http://ruleqa.spamassassin.org/?daterev=20120403-r1308758-n&rule=%2FTO_EQ_FM&srcpath=&g=Change Looks like all the hams hitting those three rules are in llanga's corpus.
(In reply to comment #1) > A new sa-update hasn't been generated since 2012-02-25, how did your scores > change a few days ago? > > Those rules are ranked pretty terribly. Best rules are ranked 1, worst are > ranked 0: > > MSECS SPAM% HAM% S/O RANK SCORE NAME WHO/AGE > 0 0.5257 0.0050 0.991 0.54 0.00 TO_EQ_FM_HTML_ONLY > 0 0.3420 0.0017 0.995 0.52 0.00 TO_EQ_FM_DIRECT_MX > 0 0.2501 0.0011 0.996 0.51 0.00 TO_EQ_FM_HTML_DIRECT > > - > http://ruleqa.spamassassin.org/?daterev=20120403-r1308758-n&rule=%2FTO_EQ_FM&srcpath=&g=Change > > Looks like all the hams hitting those three rules are in llanga's corpus. It still has a nice S/O, though. I recommend a local scoring higher until we can get masscheck chugging along!
I fear I his those because of company generated mails. I'm not sure but I think mails are constructed using some outdated mail library. Anyway this should have been reflected in my corpus earlier, shoudlnt't it ? I'm not entirely sure how to start checking on those.
(In reply to comment #3) > I fear I his those because of company generated mails. those should match trusted_networks isn't it?
Some of those rules has been removed from 72_scores.cf some time ago, time to close this bz ?
The ruleset and its scoring has evolved away from this being anything like a problem, much less a bug.