@jdemeyer, I was not absolutly clear with how the methods that have a slot interact with __dict__, I tried to read the code source and to look were __dict__ or the slot were preferred over each other and I'm still not sure about the details. I couldn't find where __getattr__ would look in the slots.

I will push a commit to update the slots instead of __dict__ tonight.

I was not absolutly clear with how the methods that have a slot interact with __dict__

It's complicated, I would have to look up the details myself.

But basically, the slots are used to put entries in the __dict__ (instances of wrapper_descriptor) and the presence of a special method (but not a wrapper_descriptor) in the __dict__ adds an entry (for example slot_nb_index) in the slot. It's especially this two-way interaction that makes things more complicated.

Thanks @jdemeyer, I think the last change should be good to make nb_int defaults to nb_index

What I meant is that should only do

type->tp_as_number->nb_int = type->tp_as_number->nb_index;

and then the wrapper descriptor __int__ should be added automatically.

This doesn't seem to work, with this:

    /* If __index__ is defined but not __int__, make it default to __index__.
       Don't touch __float__ and __complex__ as there could be some loss of
       precision.
    */
    if (type->tp_as_number != NULL && type->tp_as_number->nb_int == NULL) {
        type->tp_as_number->nb_int = type->tp_as_number->nb_index;
    }

I get:

✗ ./python.exe -m test.test_index
...E....................................................
======================================================================
ERROR: test_int_defaults_to_index (__main__.BaseTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/remi/src/cpython/Lib/test/test_index.py", line 97, in test_int_defaults_to_index
    self.assertEqual(int(Test()), 4)
TypeError: int() argument must be a string, a bytes-like object or a number, not 'Test'

Please post your failing branch, otherwise it's hard to guess what the problem is.

@jdemeyer, thanks, I pushed my last commit

#13108 has been merged so the issue is now resolved.