When record array is indexed with incorrect field name it raises ValueError as expected:

> arr = np.zeros(10, dtype=[('a', 'i8')])
> arr['b']

ValueError: no field of name b

When a simple array is indexed, it raises IndexError as unexpected:

> arr = np.arange(10)
> arr['a']

IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or 
boolean arrays are valid indices

According to this question on SO previously it was ValueError in both cases. Because you can not typecheck, they are both ndarrays, it is odd to have different types of exceptions.

Was this an intended change?

Also according to all this, the message for IndexError is not full, it does not say about fields.