Re: Re: PEAK-Rules bug

At 10:32 AM 6/25/2008 -0400, Phillip J. Eby wrote:
>At 01:08 AM 6/25/2008 +0300, Sergey Schetinin wrote:
>>This IS a bug, right?
>
>Yes.  I just got back from a trip and haven't fixed it yet.  The 
>problem is that the default "inclusion" bitmask needs to include (by 
>default) all cases based on a "not" condition.  I see one obvious 
>way to fix it, but it will then break on istype(object) (i.e., if 
>you test for something being exactly type(object), it would then be 
>true for ANY object.  So, I'm still trying to figure out how to work 
>around that.

I spoke too soon.  :)  As soon as I sent the email, I had an idea for 
how to work around that.  The fix (with expanded test coverage, too) 
is now in SVN and the snapshots directory.  I think this should be 
the last of the fallout from the istype()-support index 
refactoring.  (Unfortunately, the test coverage that sufficed for the 
old index design did not suffice for the new design, which has a much 
higher number of combinations of types of criteria, including mixed criteria.)