`_ctypes_simple_instance` in _ctypes.c returns the opposite logic of what its documentation claims. It is supposed to return true when the argument (a type object) is a direct subclass of `Simple_Type` (`_SimpleCData` in Python code). However it returns false instead.
No bugs have manifested from this because all of the call sites ( `callproc.c::GetResult`, `callbacks.c::_CallPythonObject`,_`ctypes.c::PyCData_get`, `_ctypes.c::Simple_from_outparm`) invert the return value of this function. The last example, `ctypes.c::Simple_from_outparm` only calls `Simple_get_value()` when `_ctypes_simple_instance` returns false, which makes sense because otherwise the invocation of `_ctypes.c::Simple_from_outparm()` could trigger an assertion error.
This is not just simply an issue of inverted logic because the logic isn't inverted in all cases. In `_ctypes_simple_instance` in the case when `PyCSimpleTypeObject_Check(type)` returns false, if this were supposed to be perfect inverted logic then the whole routine would return 1 (True) not 0. Fortunately, due to the way the code is structured, I don't think there is a case when `PyCSimpleTypeObject_Check(type)` returns false so the incorrect case where it returns a constant 0 is effectively dead code.
I have compiled a version of Python with the attached patch and run "make test" with no issues.
|
➜
