Cannot set value to structured data by name which dtype's shape is multi-dimension

[liyihc]

import numpy as np
import h5py
f = h5py.File('new.h5', 'a')

# single dimension
e = f.create_dataset('e', shape=10,dtype=[('a', int), ('b', float, 6)])
# works
e['b'] == np.empty((10, 6), dtype=float)
# works
e['b'] = e['b'] 

# multi dimension
d = f.create_dataset('d', shape=10,dtype=[('a', int), ('b', float, (2,3))])

# works
d[0] = np.empty(1, dtype=[('a', int), ('b', float, (2,3))])
# ValueError: When changing to a larger dtype, its size must be a divisor of the total size in bytes of the last axis of the array.
d['b'] = d['b'] 
# ValueError: When changing to a larger dtype, its size must be a divisor of the total size in bytes of the last axis of the array.
d['b'] = np.empty((10, 2, 3), dtype=float)
# ValueError: When changing to a larger dtype, its size must be a divisor of the total size in bytes of the last axis of the array.
d['b', 0] = np.empty((2, 3), dtype=float)
# ValueError: When changing to a larger dtype, its size must be a divisor of the total size in bytes of the last axis of the array.
d['b', 0] = d['b', 0]

numpy doesn't have this problem

np_d = d[()]
# works
np_d['b'] =  np.empty((10, 2, 3), dtype=float)
# works
np_d['b'] = np_d['b']

h5py    3.9.0
HDF5    1.12.2
Python  3.10.4 | packaged by conda-forge | (main, Mar 30 2022, 08:38:02) [MSC v.1916 64 bit (AMD64)]
sys.platform    win32
sys.maxsize     9223372036854775807
numpy   1.23.3
cython (built with) 0.29.35
numpy (built against) 1.21.6
HDF5 (built against) 1.12.2