I don't suppose we can actually test this @lorentzenchr 😁

I think it makes sense but I don't understand how it never triggered a segfault

I think it makes sense but I don't understand how it never triggered a segfault

IIUC, it does not really read/access the last element. Segfaults are just hard to pin down anyway.

So I tested this snippet

import numpy as np

a = np.array([1., 2, 3, 4, 5])
cdef double[::1] a_view = a
cdef double[::1] b_view = a[:-1]  # slice of a_view of all but the last element
print(b_view[4])

with boundscheck=True, cython throws an error because we try to access out of bound
with boundscheck=False, it prints 5, i.e. the last element of a. I guess it's because with boundscheck=False it just directly deref the pointer of the base array which is properly allocated and not out of bound.

In summary, the current code works but in a non-intuitive way