[Pyrex] __richcmp__ , __hash__: dependencies on subclassing ?

[Uwe Hoffmann]

	Hi,

	i have two pyrex extension types Base and Sub where Sub is
	derived from Base. Base has a __richcmp__ method but not a
	__hash__ method, Sub has a __hash__ method. Now if i compare
          instances of Sub  (a==b) it seem that this is actually a test
	for object identity (__richcmp__ is *not* called) and the result
          is always False if id(a) != id(b) and True if id(a) == id(b).
	On removing the __hash__ method of the Sub class everything is
	fine (__richcmp__ of class B is called).

	If i build the same class hierachy within native python this
	behaviour is not observed (__eq__ of the Base class is always
	called)

	Any hints ?


	regards
		Uwe


hoping that the indentation is correctly displayed

#missing.pyx
cdef class Base:

      cdef  int a

      def __new__(self,*args,**kw):
          pass

      def __init__(self,*args,**kw):
          print args[0]
          self.a=args[0]

      def __richcmp__(self,other,case):
          print "__richcmp__"
          if case == 2:
              return bool((<Base>left).a == (<Base>right).a)



cdef class Sub(Base):
      def __hash__(self):
          print "__hash__"
          return hash(self.a)



Python 2.4.2 (#2, Sep 30 2005, 22:19:27)
[GCC 4.0.2 20050808 (prerelease) (Ubuntu 4.0.1-4ubuntu8)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
  >>> from missing import Base,Sub
  >>> a=Base(1)
1
  >>> aa=Base(1)
1
  >>> b=Base(2)
2
  >>>
  >>> sa=Sub(1)
1
  >>> saa=Sub(1)
1
  >>> sb=Sub(2)
2
  >>> print a==aa
__richcmp__
True
  >>> print a==b
__richcmp__
False
  >>>
  >>> print sa==saa    # huh ?!?!       <---------
False
  >>> print sa==sb
False
  >>> print sa==sa
True
  >>>