I use the package 2.5-4.1mdv2007.1.i586 in mandriva.
I don't know if this has already been reported elsewhere but the following code:

  a= ['C','D']
  class A(object):
    __slots__ = a + ['__'+name for name in a]
    def __init(self, c, d):
      self.__C = c
      self.__D = d
    C = property(lambda self: self.__C)
    D = property(lambda self: self.__D)
  a = A()

results is an a.name attribute with value 'D'.

But using :

  b= tuple('C','D')
  class B(object):
    __slots__ = b + tuple('__'+name for name in b)
    def __init(self, c, d):
      self.__C = c
      self.__D = d
    C = property(lambda self: self.__C)
    D = property(lambda self: self.__D)
  b = B()

b appropriately do not have a b.name attribute.

Sorry, I can't follow you. Why should b have a "name" attribute?

Also, is it intentional that your __init is not named __init__?

Ah, strike my last comment.

This is not a bug, or at least not fixable, since list comprehensions do leak their iteration
variable into the enclosing scope, while generator expressions don't.

They will stop doing this in Py3k.