How does Python’s reversed() function work?

According to Python’s docs, reversed() uses __getitem__ and __len__ if __reversed__ is not implemented.

I’ve encountered a weird behavior and failed to explain it:

>>> class A(dict):
...     pass
...
>>> reversed(A())
Traceback (most recent call last):
    ...
TypeError: 'A' object is not reversible

>>> class B(dict):
...     def __getitem__(self, key):
...         return super().__getitem__(key)
...     def __len__(self):
...         return super().__len__()
...
>>> reversed(B())
Traceback (most recent call last):
    ...
TypeError: 'B' object is not reversible

>>> class C:
...     def __getitem__(self, key):
...         return "item"
...     def __len__(self):
...         return 1
...
>>> reversed(C())
<reversed object at 0x00000000022BB9B0>

Although calling reversed() on mapping types makes no sense, how does it know it’s a mapping? Does it internally check isinstance(inst, dict)? Does it check for any general mapping like collections.abc.Mapping? Is there any way to override this behavior without implementing __reversed__?

I thought it might be due to dict implementing a __reversed__ that throws a TypeError, or one that equals None much like how you disable __hash__, but dict.__reversed__ turned out empty with AttributeError thrown.

UPDATE:

New Python versions implement __reversed__ for dictionaries. Mapping protocols (such as collections.abc.Mapping) set __reversed__ to None.

Answer

Yes, there’s a check for dict type in PySequence_Check used by reversed.

// cpython/Objects/enumobject.c
if (!PySequence_Check(seq)) {
    PyErr_Format(PyExc_TypeError,
                 "'%.200s' object is not reversible",
                 Py_TYPE(seq)->tp_name);
    return NULL;
}


// cpython/Objects/abstract.c

int
PySequence_Check(PyObject *s)
{
    if (PyDict_Check(s))
        return 0;
    return s != NULL && s->ob_type->tp_as_sequence &&
        s->ob_type->tp_as_sequence->sq_item != NULL;
}