setter on metaclass is not invoked

[slozier]

Ran into this issue while trying to get the typing module working:

class MetaClass(type):
    def __new__(metacls, name, bases, attrdict, **kwargs):
        return type.__new__(metacls, name, bases, attrdict)

    @property
    def test(self):
        return 1

    @test.setter
    def test(self, value):
        raise Exception

class Generic(metaclass=MetaClass):
    pass

assert Generic.test == 1

try:
    Generic.test = 2
except:
    pass
else:
    assert False

@BCSharp since you looked at metaclasses a bit, any ideas?

[BCSharp]

@slozier I did look at metaclasses a bit and a bit more at the core language part, but somehow handling of decorators never came up as an issue before so I have no clue what is going on here.

[slozier]

@BCSharp Sounds good, I'll dig into it a bit more I guess.

[BCSharp]

@slozier This may be not related to the problem at hand but just for the record (and just in case it is), the metaclass support is not yet complete, esp. how the classdict is being handled. I believe that the classdict should be able to be any mapping, to match CPython and eliminate all kinds of small corner incompatibilities. This would be my rough plan to this end:

1. Remove __classcell__ or at least do not use classdict to pass it around (doesn't work well with metaclass __prepare__ anyway, which may even not support deletion of attributes).
2. Replace all class variables with class attributes, except __class__.
3. Change the class lambda to take its context as parameter, rather than the parent context as it is now. This will require moving of class context creation to outside of the class lambda.
4. Support arbitrary attribute dict type as classdict.
5. Complete implementation of __prepare__. At this point, PEP 3115 implementation in WhatsNewInPython30.md can be checked off.

Unfortunately, it doesn't seem I will have much free time for IronPython in the very near future, so feel free to advance any of it, if you feel inspired.

[slozier]

Unfortunately, it doesn't seem I will have much free time for IronPython in the very near future, so feel free to advance any of it, if you feel inspired.

@BCSharp That's too bad (for me at least!). Hope you're doing other interesting things. Thanks for outlining your plan (although it seems unlikely I'll get around to it in the near future, right now I'm just focusing on testing the alpha with common 3rd party packages and seeing what issues I run into).