BUG: np.rec.fromarrays(...) may fail if a field contains arrays of the same shape[0]

[herrlich10]

Describe the issue:

np.rec.fromarrays(...) may fail if a field contains arrays of the same shape[0]. The error will not occur if the arrays are of different shape[0], e.g., [np.zeros(2), np.ones(3)]. The error will not occur if written as np.array([(np.zeros(2), 'a'), (np.ones(2), 'b')], dtype=[('field1', object), ('field2', '<U1')]).

Reproduce the code example:

import numpy as np
np.rec.fromarrays([[np.zeros(2), np.ones(2)], ['a', 'b']], names=['field1', 'field2'])

Error message:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-1-eecf8753473e> in <module>
      1 import numpy as np
----> 2 np.rec.fromarrays([[np.zeros(2), np.ones(2)], ['a', 'b']], names=['field1', 'field2'])

~/PythonPlus/anaconda3/lib/python3.9/site-packages/numpy/core/records.py in fromarrays(arrayList, dtype, shape, formats, names, titles, aligned, byteorder)
    679         testshape = obj.shape[:obj.ndim - nn]
    680         if testshape != shape:
--> 681             raise ValueError("array-shape mismatch in array %d" % k)
    682 
    683     _array = recarray(shape, descr)

ValueError: array-shape mismatch in array 1

NumPy/Python version information:

1.20.3 3.9.7 (default, Sep 16 2021, 08:50:36)
[Clang 10.0.0 ]

Context for the issue:

No response

[seberg]

Note that unlike 1.20.x the call also fails if the shapes of the arrays do not match on newer versions. So I suspect if you want the object field here, more you have to expect to do extra steps to get it.
I am unsure whether using shape and dtype or similar shouldn't work though, so I suspect there are enhancements possible, I just suspect the error in itself is OK, since the discrepancy is gone.