why does numpy.trapz behave differently with booleans than with ints?

Question

Booleans in Python are implemented as ints and usually there is no difference in numpy and other packages when changing 0 by False and 1 by True. Why then these two computations differ?

>>> import numpy as np
>>> np.trapz([1,1])
1.0
>>> np.trapz([True, True])
0.5

The same thing happens with scipy.integrate.

Answer 1

Found it. It seems to be a bug: https://github.com/numpy/numpy/issues/12858

As pointed in the issue, the key is that

>>> True + True
2
>>> np.True_ + np.True_
True