Wrong matrix indicies for mesh generated by Sphere


When I set up a matrix for a surface mesh generated by Sphere with different ansatz and test spaces (e.g. P0 and P1) and then extract the matrix entries, the indices are wrong. See the example in
testIJC.edp (912 Bytes)
to reproduce the situation. In particular, the indices in I range from 0 to the size of the martix K (instead of K.n-1).

Is this a bug, or do I something wrong here?