In the current upstream version, two unit tests fail, but I don't see why. Their error is only slightly above the threshold.
On a related matter: the unit tests compare the HDF files looking at their data - this is of course correct.
however, if we add the comparison of "iteration" counter (saved in HDF5 files as well), then we can see if the code did not perform the exact same number of time steps. I think that would be quite helpful. it's an integer - this is even an exact comparison.
if the difference between two files is small and the iteration counter is different, then we likely have found the origin of the problem.
could you implement that?
In the current upstream version, two unit tests fail, but I don't see why. Their error is only slightly above the threshold.
On a related matter: the unit tests compare the HDF files looking at their data - this is of course correct.
however, if we add the comparison of "iteration" counter (saved in HDF5 files as well), then we can see if the code did not perform the exact same number of time steps. I think that would be quite helpful. it's an integer - this is even an exact comparison.
if the difference between two files is small and the iteration counter is different, then we likely have found the origin of the problem.
could you implement that?