specify bcondition at single vertex in compute_rotated

lapy/diffgeo.py

@@ -366,7 +366,11 @@ def tria_compute_rotated_f(tria, vfunc):
     #   we can pass identity instead of B here:
     # div is the integrated divergence (so it is already B*div)
     fem.mass = sparse.eye(fem.stiffness.shape[0], dtype=vfunc.dtype)
-    vf = fem.poisson(divf)
+    # since the solution is ill defined (addition of constant)
+    # we specify an arbitary boundary condition at a single vertex to
+    # remove that degree of freedom
+    dtup = ( np.array([0]), np.array([0.0]))
+    vf = fem.poisson(divf,dtup)
     return vf