mpl 2.0rc2, numpy 1.11.3
An invalid value, whether masked or not, leads to two RuntimeWarnings:
imshow(np.ma.masked_invalid([[1, 2], [3, np.nan]]))
and
imshow([[1, 2], [3, np.nan]])
give
/usr/lib/python3.5/site-packages/matplotlib/image.py:371: RuntimeWarning: invalid value encountered in less
rgba[..., 1] = A < 0 # under data
/usr/lib/python3.5/site-packages/matplotlib/image.py:372: RuntimeWarning: invalid value encountered in greater
rgba[..., 2] = A > 1 # over data
I would argue that a warning should be raised in neither case, given that such inputs are supposed to be supported (http://matplotlib.org/examples/pylab_examples/image_masked.html, #6069).