GeezerVison

No. The isinstance check only ran when the value was not already a ctypes object, and a true result was still cast to CDataObject so its b_ptr could be used. There is no valid object for which that cast works.

A real ctypes value never took that branch. CDataObject_Check uses the real type, and instances of c_int (including a Python subclass) are subclasses of _ctypes._CData. Assigning one still goes through the same path as before:

class MyInt(ctypes.c_int):
    pass

ptr = ctypes.pointer(ctypes.c_int(1))
ptr.contents = MyInt(7)  # still accepted

The objects that pass isinstance and fail CDataObject_Check are objects whose __class__ is not their real type. Using those as a CDataObject crashed, which is the bug in the report. They now raise TypeError.

Assigning a different real ctypes type, such as c_char to a pointer to c_int, is unchanged. That was already accepted without the isinstance check.