Speed up MRO calculationa93755ddba9a176

Optimize gi.type.mro() with the following observations and tricks:

1. Python prepares all the base classes before trying to calculate the
   MRO of the current one (it first needs to populate __bases__, for
   example). So we can assume that the base class MRO is already available
   in __mro__ and this will have been previously calculated (by us, in the
   case of gi classes). This avoids repeating a lot of MRO-calculating work,
   and also avoids (re)calculating MROs for inheritance chains that don't
   have any gi classes in them anyway.

2. With that simplification in place, we can avoid recursion, which is not
   all that great in Python...

3. ...except in the uncommon case of a Python2 old-style classes, where
   __mro__ is not available. There doesn't seem to be any existing
   function to calculate or read MRO of old-style python classes, so just
   keep doing as before: calculate the C3 MRO of the old-style class via
   recursion. That behaviour is not really correct, and the recursion is
   not desirable, so we print a warning here.

This makes the "hello world" Sugar app start up approximately 0.5 seconds
faster on XO-1.5.

https://bugzilla.gnome.org/show_bug.cgi?id=703829

0 files changed, 0 insertions, 0 deletions