Dear Luca,
Thank you very much for your reply, your answer was very precise and,
therefore, I could solve my problem. Effectively, I was not considering the
subfaces when I was assembling, and of course the "get_dof_indices"
function was not recognizing the nodes neighbour nodes those cells.
Dear Juan,
The logic you should use is: when your neighbour is finer, you should not
assemble the full face term from the coarse neighbour. Instead, you have to
assemble your DG terms using a FESubfaceValues initialized with the correct
arguments on the coarse face, to match *each* of the fine