I believe that the fix to Ticket #808 has caused a regression in numpy.insert()

This is the commit

Previously:

np.insert([[1,1,1]], 0, [2,2,2], axis=0)

gave:

 array([[2, 2, 2],
      [1, 1, 1]])

now it gives

array([[2, 2, 2],
   [2, 2, 2],
   [2, 2, 2],
   [1, 1, 1]])

I don't think is the correct result.

Also issue #378 is a result of same change