As of now we expect most code to be triggered by the translate test because we have carefully hand checked. Going forward and because there is code difference between what GEOS and SHiELD wants to run for the dycore (among other challenges) we need to have a system that make sure we cover all of our ported code with the translate tests.
This won't be 100% fault-proof since we still could run the code with inputs that don't trigger side-cases in the numerics that create issues, but it will help making sure we have a reasonable coverage, especially when regenerating translate test data.
As of now we expect most code to be triggered by the translate test because we have carefully hand checked. Going forward and because there is code difference between what GEOS and SHiELD wants to run for the dycore (among other challenges) we need to have a system that make sure we cover all of our ported code with the translate tests.
This won't be 100% fault-proof since we still could run the code with inputs that don't trigger side-cases in the numerics that create issues, but it will help making sure we have a reasonable coverage, especially when regenerating translate test data.
Parent: https://github.com/GEOS-ESM/SMT-Nebulae/issues/41
pyFV3
when running the full set of tests