tests in 'master' fail probably due to sensitivity of adaptive grid
For employing the linear solver, I changed the FEM in the GridFunctionSpace of FEM order 0 to P0. However, this GFS is also used for the Adaptivity GFS and the change in FEM apparently makes a difference in the refinement algorithm.
Apparently, the changes introduced by !3 (merged) increased the sensibility of the grid adaptation algorithm to a level where solutions differ depending on system and setup. Propositions to solve this issue:
- generally soften tolerances
- remove refence cases with adaptive grids