Auto-promote behavior? Re: svn commit: r881199 - /spamassassin/trunk/rules/active.list

[Warren Togami <wt...@redhat.com>]

On 11/17/2009 03:50 AM, jm@apache.org wrote:
>
>   # good enough
> -VANITY
> -

For a while this rule was auto-promoted and pushed to the 3.3.0 
sa-update channel.  But November 17th's promotions validation mail says 
VANITY was removed from "good enough".  I see no FP's in the nightly 
masschecks of the previous week that could cause this.  November 14th's 
masscheck is the last day where it says VANITY instead of T_VANITY in 
ruleqa.

This leaves me confused as to how auto-promote is supposed to behave?

Warren Togami
wtogami@redhat.com

Re: Auto-promote behavior? Re: svn commit: r881199 - /spamassassin/trunk/rules/active.list

["Daryl C. W. O'Shea" <sp...@dostech.ca>]

On 24/11/2009 12:12 AM, Warren Togami wrote:
> http://www.gossamer-threads.com/lists/spamassassin/commits/144427
> Yesterday's auto-promotions brought back many of the rules that were
> demoted last week for seemingly no reason.  But the VANITY rule remains
> demoted.  Any idea what is going on here?

I've always felt that there was too much bounce happening in this
process.  I don't think the script collects any info about the previous
state of the list (ie I think it's based on single numbers, not a high
number to get on the list and a lower number to be kicked off the list).

I would suspect that there's another criteria that isn't being met...
perhaps the S/O is fine but the percentage of mail hit isn't always high
enough.

Daryl

Re: Auto-promote behavior? Re: svn commit: r881199 - /spamassassin/trunk/rules/active.list

[Warren Togami <wt...@redhat.com>]

On 11/22/2009 03:29 AM, Warren Togami wrote:
> On 11/22/2009 01:32 AM, Warren Togami wrote:
>> On 11/17/2009 03:50 AM, jm@apache.org wrote:
>>>
>>> # good enough
>>> -VANITY
>>> -
>>
>> For a while this rule was auto-promoted and pushed to the 3.3.0
>> sa-update channel. But November 17th's promotions validation mail says
>> VANITY was removed from "good enough". I see no FP's in the nightly
>> masschecks of the previous week that could cause this. November 14th's
>> masscheck is the last day where it says VANITY instead of T_VANITY in
>> ruleqa.
>>
>> This leaves me confused as to how auto-promote is supposed to behave?
>>
>
> Something appears to have gone haywire.
>
> http://ruleqa.spamassassin.org/20091121-r882858-n/T_RCVD_IN_PSBL/detail
>
> How did RCVD_IN_PSBL become deactivated?
>
> Warren

http://www.gossamer-threads.com/lists/spamassassin/commits/144427
Yesterday's auto-promotions brought back many of the rules that were 
demoted last week for seemingly no reason.  But the VANITY rule remains 
demoted.  Any idea what is going on here?

Warren

Re: Auto-promote behavior? Re: svn commit: r881199 - /spamassassin/trunk/rules/active.list

[Warren Togami <wt...@redhat.com>]

On 11/22/2009 01:32 AM, Warren Togami wrote:
> On 11/17/2009 03:50 AM, jm@apache.org wrote:
>>
>> # good enough
>> -VANITY
>> -
>
> For a while this rule was auto-promoted and pushed to the 3.3.0
> sa-update channel. But November 17th's promotions validation mail says
> VANITY was removed from "good enough". I see no FP's in the nightly
> masschecks of the previous week that could cause this. November 14th's
> masscheck is the last day where it says VANITY instead of T_VANITY in
> ruleqa.
>
> This leaves me confused as to how auto-promote is supposed to behave?
>

Something appears to have gone haywire.

http://ruleqa.spamassassin.org/20091121-r882858-n/T_RCVD_IN_PSBL/detail

How did RCVD_IN_PSBL become deactivated?

Warren

Auto-promote score?

[Warren Togami <wt...@redhat.com>]

spamassassin/trunk/rulesrc/sandbox/khopesh/20_khop_bl.cf

meta KHOP_GREYED __GREYLISTED && (RDNS_NONE || RDNS_DYNAMIC || 
__HELO_NO_DOMAIN)
describe KHOP_GREYED    Greylisted and sent from dynamically-named relay
score    KHOP_GREYED    0.1

Say for example the masscheck shows good statistics on this (or any) 
rule and the rule becomes auto-promoted.  In the past it seems 
auto-promoted rules are included in the sa-update channel without a 
score line.  This means they get 1 point regardless of what was intended 
in the sandbox.

Was this intended behavior?

Is this an auto-promote bug?

Warren