Replying myself to my own message... The differences between the old and new PW/compute_dip.f90 (see function saw()) actually explain my apparently weird results for 4.3.1, which occur when there is electron charge in the region where the electric field is reverted.
In the previous version 4.0.1 this was not a problem for the calculation of the dipole itself, but the application of such an electric field is probably incorrect in most cases of physical interests. The two lead to the same result for a well-constructed cell setup. Guido Fratesi -- Guido Fratesi Dipartimento di Scienza dei Materiali Universita` degli Studi di Milano-Bicocca Milano, Italy
