Clean up all the places where role processing occurs. This is now in a central place in hyperdb.Class and is used consistently throughout. This also means now a template can override the way role processing occurs (e.g. for elaborate permission schemes). Thanks to intevation for funding the change. Note: On first glance the hyperdb.Class may not be the ideal place for role processing. On second thought: Roles may appear in other classes, too (e.g., a user_group or similar) which then don't need to reinvent the wheel. And I didn't want to introduce a separate UserClass (as is the case for the HTML classes) due to compatibility issues with existing schema.py out there.

--- a/CHANGES.txt	Mon Dec 14 21:48:51 2009 +0000
+++ b/CHANGES.txt	Tue Dec 15 15:11:27 2009 +0000
@@ -41,6 +41,11 @@
   would trigger a traceback about an unbound variable.
   Add new regression test for this case. May be related to (now closed)
   issue1177477. Thanks to Intevation for funding the fix.
+- Clean up all the places where role processing occurs. This is now in a
+  central place in hyperdb.Class and is used consistently throughout.
+  This also means now a template can override the way role processing
+  occurs (e.g. for elaborate permission schemes). Thanks to intevation
+  for funding the change.
 
 2009-10-09 1.4.10 (r4374)