Message3293

this is fixed on the asm branch with:

r4684, r4693, r4694

Note that the iadd.py test still doesn't work exactly as CPython does, 
it's now:

123
type(l) is NotList False
type(l_too) is list True

Which matches pypy. The discrepancy is due to CPython's odd behavior 
with builtins

Basically in this example, list __iadd__ works on iterable objects:

Python 2.5.2 (r252:60911, Apr 22 2008, 12:00:45) 
[GCC 4.0.1 (Apple Computer, Inc. build 5367)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from iadd import NotList
>>> l = []
>>> l.__iadd__(NotList([1,2,3]))
[1, 2, 3]

Since NotList is iterable, why is NotList.__radd__ called instead of 
list.__iadd__ using NotList.__iter__? Because CPython makes subclass 
slots (like __add/radd__) take precedent over builtin slots in cases 
like this. A better example is:

Python 2.5.2 (r252:60911, Apr 22 2008, 12:00:45) 
[GCC 4.0.1 (Apple Computer, Inc. build 5367)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> class Base(object):
...     def __imul__(self, other):
...             return 'Base imul'
... 
>>> class Foo(Base):
...     def __mul__(self, other):
...             return 'Foo mul'
... 
>>> f = Foo()
>>> f *= 1
>>> f # super __imul__ was consulted before Foo __mul__
'Base imul'
>>> class Bar(list):
...     def __mul__(self, other):
...             return 'Bar mul'
... 
>>> b = Bar()
>>> b *= 1
>>> b # super __imul__ *NOT* consulted before Bar __mul__
'Bar mul'
>>> list.__imul__ # there's really a super __imul__ slot
<slot wrapper '__imul__' of 'list' objects>

Whereas Jython/Pypy would have called list.__imul__ and returned an 
empty list

CPython's behavior seems inconsistent, and I don't see it documented 
anywhere (reading http://docs.python.org/ref/coercion-rules.html gives 
you the impression Jython/Pypy are correct).

We and the Pypy folk probably need to follow up on this with CPython 
folk. (ccing pedronis)