The following code (simplified from a real example) triggers an unexpected TypeError:

    @dataclass
    class PrioritizedItem:
        priority: 'int'
        def __eq__(self, other):
            return self.priority == other.priority

This code gives the following trackback:

    Traceback (most recent call last):
      File "/Users/raymond/Documents/tmp16.py", line 16, in <module>
        @dataclass
      File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/dataclasses.py", line 598, in dataclass
        return wrap(_cls)
      File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/dataclasses.py", line 590, in wrap
        return _process_class(cls, repr, eq, order, hash, init, frozen)
      File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/dataclasses.py", line 539, in _process_class
        _set_attribute(cls, '__hash__', None)
      File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/dataclasses.py", line 444, in _set_attribute
        raise TypeError(f'Cannot overwrite attribute {name} '
    TypeError: Cannot overwrite attribute __hash__ in PrioritizedItem

I think the error message here can be improved to explain why is being raised.  Otherwise the cause isn't self-evident.

Thanks.

I'm going to be changing the errors as part of #32513, but I'll make sure to cover this case better.

No matter what happens with #32513, I think the right thing to do in this case is to not error if trying to set __hash__ to None, if it already is None in __dict__.

This has been fixed as part of #32513.