Thorough check that ALL tests use epsilons for floating point comparison in boundary conditions and spatial params
This is the main reason for failing test. All tests should use a proper epsilon.
- This epsilon usually is made of a base epsilon
static constexpr eps_ = 1.5e-7;
times a magnitude. In the case of grids this should be a characteristic cell size. - If the grid is in the scale of 1m there is no adjustment necessary. However a runtime check would make it flexible for using any grid
- We could also introduce a preprocessor constant or global variable in a namespace for the epsilon
- We can also use more of Dune::Float_Cmp