This is related to [python-Bugs-1648268], but I think these problems might be important enough to consider fixing prior to any patch being produced for that item.

Modules/datetimemodule.c:time_isoformat() is declared in time_methods[] as METH_KEYWORDS. However, I believe it is better declared as METH_NOARGS (calling it with args and kwargs doesn't raise any exception, but it doesn't accept them).

Objects/typeobject.c:4428 - slot_nb_inplace_power is declared with the SLOT1() macro. I'm not sure I entirely grok what's going on here (not enough to supply a python-level failure case), but it seems to me that it should be declared with the SLOT2() macro (it's a ternary op). FWIW, I changed it from:

SLOT1(slot_nb_inplace_power, "__ipow__", PyObject *, "O")

to:

SLOT2(slot_nb_inplace_power, "__ipow__", PyObject *, PyObject *, "OO")

... and that ran the failing tests OK.

Hopefully someone familiar with this code can determine if this is correct or not.

Thanks, Kev.

Thanks for the report. I have fixed the first bug, in r53671 and r53672.

As for the second bug: I think your suggested change is wrong. __ipow__ is supposed to take only two arguments. I'm unsure why nb_inplace_power is defined with three arguments; the third argument is set to Py_None in all places I could find. So it is, at a minimum,
ok if slot_nb_inplace_power discards its third argument; I wonder whether the API should be changed to drop this argument entirely. This is for python-dev to discuss.

The second bug should now be fixed in r53689 and r53690.

I had to revert r53672. New fix is r53816.