Juat have to wait and see what next weekend rescore does for net. I ran many extra runs last time fixing things, who knows what happened..
On Wed, Jul 29, 2020 at 02:52:17PM -0400, Kevin A. McGrail wrote: > Henrik, might be something broken on the new system. > > > > -------- Forwarded Message -------- > Subject: Re: Mailspike rules all return 0.0 > Date: Wed, 29 Jul 2020 12:38:37 -0400 > From: Bill Cole <[email protected]> > Reply-To: [email protected] > To: [email protected] > > > > On 29 Jul 2020, at 9:27, Simon Harwood wrote: > > > Hello, > > > > I have noticed that the mailspike rules are enabled in SpamAssasin but > > all return zero values: > > > > 0.0 RCVD_IN_MSPIKE_BL Mailspike blacklisted > > 0.0 RCVD_IN_MSPIKE_L5 RBL: Very bad reputation (-5) > > This is actually a rounding issue: all RCVD_IN_MSPIKE_* rules are being > scored at +/- 0.001 by RuleQA so the score report rounds it to 0.0. They > are rescored that way because they are not hitting anything in the > submitted corpora but have scores set in the static scores list. > > That MAY indicate a bug in RuleQA, since I don't believe that "net" > rules like these should be rescored by RuleQA. Perhaps someone more > familiar with the mechanics of RuleQA will correct me... > > > > Please can you tell me where in the SpamAssassin configuration I find > > the setting(s) that need to be changed such that non-zero values are > > returned. > > I have been receiving large quantities of spam that instead of just > > being flagged as spam would have met the conditions for being rejected > > completely if either of the mailspike values had been 0.1 or greater. > > As Claudio said, local.cf is the place. Where your local.cf should be > depends on how your SpamAssassin was built, but there should be a > baseline version in the site rules directory. You can find that with: > > spamassassin --lint -D config 2>&1 |grep 'site rules dir' > > > -- > Bill Cole > [email protected] or [email protected] > (AKA @grumpybozo and many *@billmail.scconsult.com addresses) > Not For Hire (currently) >
