https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6510

--- Comment #5 from John Hardin <[email protected]> 2010-11-06 11:53:00 UTC ---
(In reply to comment #4)
> (In reply to comment #1)
> > +  sed -e 's/0\.000/0.001/g' < trunk-rulesrc-scores/72_scores.cf >
> > trunk/rules/72_scores.cf
> 
> Yeah, that won't work well.  It'll enable net rules in non-net scoresets.

Would that cause problems given the presence of "tflags net" on net rules?
Surely SA does not rely on the score being zero to disable net rules when run
in local mode...

> I think I'll change the script to abort the update if lint detects dependency
> issues.  I missed tonight's update (I think there's yet another rule or more
> that will have zeroing issues)... so I'll look at it tomorrow.

That would keep the update from breaking things at the cost of possibly
blocking rule and score updates for extended periods while the rescorer is
misbehaving.

If we can put a check into the rescorer that does not allow scores to go to
zero on published rules (taking into account net vs. non-net as my hack does
not, if that's indeed critical) then that should be done. As I said earlier, if
a rule is performing well enough to be published then it should not be assigned
a zero score.

Any idea why the scores are jumping around so much? That bothers me too, though
it's not as problematic as scores going to zero for no discernible reason.

-- 
Configure bugmail: 
https://issues.apache.org/SpamAssassin/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

Reply via email to