BUG: Masked scalar comparison returns float

>>> (numpy.ma.array(5.5, mask=True) > 0).dtype
dtype('float64')

Moreover, the result has float dtype even if the masked scalar is an int:

>>> (numpy.ma.array(5, dtype='i', mask=True) < 0).dtype
dtype('float64')

The problem is not present in oldnumeric:

>>> (numpy.oldnumeric.ma.array(5.5, mask=True) < 0).dtype
dtype('bool')

>>> numpy.__version__
'1.8.0'

Provide feedback