Many unit tests are failing, many due (I think) to updates to numpy that have changed how it deals with arrays that contain lists of different lengths. Attractors can still be computed when there are only fixed points, but cyclic attractors lead to errors.
Also things like np.inf -> inf.