The main build system is currently make. However, it is very fragile and would need some refactoring. The issue is that definitions and declaration are not well seperated. In my development branch, one can find a working cmake build system.
One problem is the memory efficient storage the generalized Madelung matrix, which is needed to solve the electrostatic problem. The generalized Madelung matrix is hypermatrix with the two lattice site dimensions, one l angular momentum and one m magnetic quantum number dimension. At the moment, this tensor is stored completely as four dimensional matrix. A transferable data structure for this problem would enable even non-spherical treatment of the potential.