http://issues.apache.org/SpamAssassin/show_bug.cgi?id=4347





------- Additional Comments From [EMAIL PROTECTED]  2006-07-12 01:31 -------
> Please vote on what to do and it's corrosponding patch.
> * 3448: lint error when meta dependencies are missing/disabled
> * 3574: provide debug info for missing/disabled meta dependencies
> * do nothing

I'm still in favor of a warning on a meta for the following cases:
a. Referenced rule name does not exist (ie: mis-spelled in meta, or not 
available in the current rules.
b. Meta depends on a disabled plugin.

I am NOT in favor of a warning in the following case:
c. Meta depends on a rule with zero score.  (But a debug message would be nice.)

Logic behind these suggestions:

Item A is a frequent problem in SARE masschecks.  I've been burned myself any 
number of times.  While griping on the user's list doesn't show up, it is 
frequent and loud on the SARE list where we frequently build metas.

There are really two problems here: mis-spelling a term in the meta, and 
forgetting to include all of the necessary required rules to run a masscheck 
(which runs with ONLY the submitted rules, not the standard rules.)  Both of 
those reasons end up causing more than their fair share to pulled hair and 
wasted masscheck runs.  A warning would give a strong hint of the problem.

In item B, if the meta depends on a disabled plugin, the meta should be in an 
ifplugin group itself.  After all, it ain't never gonna work right without the 
plugin, so there is no point in having it sitting there doing nothing but 
wasting space and time.  (Yea, one COULD make a meta that depended in OR 
condition on 5 different plugins, with the hope of making a score if any of 
them happen to be enabled.  I have a simple solution for that case: don't.)

For item C, we don't warn now for zero-score rules, because a zero score is how 
you disable a rule.  It follows that this will disable a meta based on that 
rule (again, unless in an OR condition with something still alive).  Since we 
don't warn for zeroed rules, we shouldn't warn for effectively zeroed metas 
that are dependent, either.  However, a debug message would be worthwhile IF 
the meta score is not itself also zeroed.

General comment on the concept of warning on a non-existant dependency: you do 
this right now for scores and descriptions (which are dependencies of the 
actual rule) if the actual rule is missing.  I don't see a meta as being any 
different than a description in this case.  Either there should be a warning 
for a missing dependency on a meta, or the warning for missing rule in 
descriptions and scores should be removed.  Obviously I don't vote for removing 
the existing warnings.  But it is worth considering, if it is decided that a 
meta shouldn't warn for missing dependency.




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to