homepage

Why use n - copy.copy(m)? It's much simpler to use n = m[:].
Note that dospath probably also has this. And macpath should have one.
I'm +1 on fixing the functionality if it's indeed broken as Skip reports; haven't looked at the code in detail (except I noted the import copy).

Check it in -- you already checked in the changes to the test!
I propose to use m[:] instead of copy.copy(m), and not import the copy module.

splitting on os.sep isn't good enough for Windows (it supports both "/" and "\"). the correct way would be to split on either os.sep or os.altsep (if not None), or to normalize the path's before comparing them.
except for this, I'm +0 on this one.

Guido asked:
 > Why use n - copy.copy(m)? It's much simpler to use n = m[:].
Just a brain fart. I will upload a revised patch.
I will also try splitting on both os.sep and os.altsep as Fredrik indicated.

This version of the patch tries splitting with os.sep then
if that doesn't create a length > 1 list tries os.altsep.
It also adds the same code to ntpath.py, dospath.py and macpath.py. Someone needs to test them!

checked in