[PEAK] Possible bug in RuleDispatch?

[Phillip J. Eby]

At 06:39 PM 10/5/2006 +0200, Alberto Valverde wrote:
>Hi,
>
>After chasing a bug in a custom dispatcher I'm writing I've stomped
>into a possible bug in RuleDispatch when dealing with integer ranges.
>
>Attached is a test-case which describes the problem better.

Thanks; it's fixed in SVN now.  The problem was that when a dispatch table 
returns None, the dispatch logic was using a default dispatch node that had 
the *class* NoApplicableMethods in it, instead of an instance.  So, fixing 
it was just adding a pair of parentheses to two files.  :)

(Just an FYI, but this is the sort of bugs one gets by not doing TDD; 
RuleDispatch's test suite has never been remotely close to adequate for 
very basic parts of the system like this.  Which is another reason why I 
will be replacing it with PEAK-Rules, which is being written thoroughly 
"test first".)