"data type not understood" with readonly array argument

[batterseapower]

Reporting a bug

The following program should succeed, because astype makes a copy of its argument:

@nb.jit()
def numba_astype(xs):
    ys = xs.astype(nb.float32)
    ys[0] = 1.0
    return ys

xs = np.array([1.0], dtype='f4')
xs.flags.writeable = False
numba_astype(xs)

In fact it just throws TypeError: data type not understood. (This might be related to the fact that the type of the array returned by astype is still marked readonly if the inital array is readonly - i.e. the readonly flag is not being cleared.)

[stuartarchibald]

Thanks for the report. I can reproduce. Fixed in #3845

[batterseapower]

That was impressively fast. Thanks!

[stuartarchibald]

No problem!