Unexpected behavior when passing sliced array to ufuncify function

[h3jia]

Hi, please see the following code:

import sympy as sp
import numpy as np
from sympy.utilities.autowrap import ufuncify

a = sp.Symbol('a', real=True)
b = sp.Symbol('b', real=True)
foo = ufuncify(args=[a, b], expr=a+b, backend='cython')

def no_copy(fun):
    
    def fun_n(mm):
        aa = mm[:, 0]
        bb = mm[:, 1]
        rr = fun(aa, bb)
        return rr
    
    return fun_n

print(no_copy(foo)(np.array([[0.,1.],[2.,3.],[4.,5.]])))
# returns [1. 3. 5.]

def with_copy(fun):
    
    def fun_w(mm):
        aa = np.copy(mm[:, 0])
        bb = np.copy(mm[:, 1])
        rr = fun(aa, bb)
        return rr
    
    return fun_w

print(with_copy(foo)(np.array([[0.,1.],[2.,3.],[4.,5.]])))
# returns [1. 5. 9.]

foo is simply adding a and b, while no_copy and with_copy aim to take in a matrix and add the elements of the first two columns in each row, but only with_copy gives the expected result.

I'm not familiar with how ufuncify works. It seems that without np.copy, ufuncify does not pass a sliced array to cython correctly?