Inconsistency using NpT and space.periodic_general

[sebasb03]

Hi all!
We are a bit unclear about using simulate.npt_nose_hoover together with space.periodic_general.

• In the examples it is stated that fractional_coordinates=True is required for NPT.
• According to the docstring and the code itself, the shift_fn returned by space.periodic_general using fractional_coordinates=True expects R to live in the unit cube and dR to live in Cartesian coordinates.
• The forces used to propagate the velocities in exp_iL2 and finally to generate dR in exp_iL1 are, however, obtained as the gradient of the energy w.r.t. the positions in the unit cube.
• We thus are currently under the impression that the forces are expressed in an incorrect system of coordinates and, in particular, in incorrect units.

Are we misunderstanding the implementation of the shift function or have we overlooked a transformation taking care of this?

Thanks for taking a look & kind regards!