Testing

My testing for now is only to run the base_mesh part of each test case on my laptop. I am not checking if they are bit-for-bit (in fact, with the change in Earth radius that this PR introduces, I don't expect them to be bit-for-bit).

A check mark in the list above indicates that base_mesh ran successfully.

Testing on Grizzly

Verification that results are bit-for-bit in base_mesh, culled_mesh and initial_state steps (Note: some tests only have 1 or 2 of these):

• Gaussian_hump/USDEQU120cr10rr2/build_mesh/
• coastal/Maine/init
• coastal/USDEQU120cr10rr2/build_mesh
• global_ocean/ARM60to10/init
• global_ocean/CA120to3/build_mesh
• global_ocean/EC60to30/init
• global_ocean/EC60to30wISC/init
• global_ocean/HI120to12/build_mesh
• global_ocean/QU240/init
• global_ocean/QU240wISC/init
• global_ocean/SO60to10wISC/init
• global_ocean/WC12/init
• global_ocean/WC14/init
• hurricane/USDEQU120at30cr10rr2/build_mesh
• hurricane/USDEQU120at30cr10rr2WD/build_mesh
• hurricane/USDEQU120at30cr10rr2WD_veg/build_mesh
• hurricane/USDEQU240at60cr20rr4/build_mesh
• hurricane/USDEQU240at60cr20rr4WD/build_mesh
• hurricane/USDEQU240at60cr20rr4WD_veg/build_mesh
• hurricane/USDEQU60at15cr5rr1/build_mesh
• hurricane/USDEQU60at15cr5rr1WD/build_mesh
• hurricane/USDEQU60at15cr5rr1WD_veg/build_mesh
• hurricane/USDEQU60at15cr5rr500WD/build_mesh
• tides/USDEQU120at30cr10/build_mesh

@mark-petersen and @sbrus89, I have a test compass environment, test_compass_0.1.10 on Grizzly for testing this branch and MPAS-Dev/MPAS-Tools#314. The procedure would be very similar to what we have done for #563 and the same tests are likely appropriate.

This new environment includes some new geometric features so you will not be able to sue your old tests as a baseline. (These new features are included in the release version of compass_0.1.9, though they were not in the test version test_compass_0.1.9 that we used yesterday.)

@mark-petersen and @sbrus89, I'm starting to test this now. Just a heads up.

@mark-petersen and @sbrus89, I realized that my tests for this PR are non-bit-for-bit because I changed the Earth radius in MPAS-Dev/MPAS-Tools#314. I'm going to pull that change out of MPAS-Dev/MPAS-Tools#314 for now and make a new test environment without that change. Then, we can have a separate PR to MPAS-Tools just for that non-bit-for-bit change before we do the next mpas_tools/compass release. Again, just a heads up.

I would like to keep ocean/coastal off limits to adding new test cases that use mpas_tools.mesh.creation until this goes in. @sbrus89 and @proteanplanet, what are the plans for the remaining development on ocean/coastal before you stop using it entirely? If they don't relate to this PR, you can hard-reset ocean/coastal to ocean/develop and proceed as far as I'm concerned.

@sbrus89 and @mark-petersen, all my testing has now passed bit-for-bit. I tested all the cases listed above through initial_state (or the last step they have if there isn't an initial_state step).

I also tested forward mode on the QU240 and QU240wISC runs, comparing the current ocean/develop against a test merge of this branch into ocean/develop and those were also bit-for-bit.

Getting this tested and merged is somewhat of a priority because it is holding up #668, #669 and #671. Also, I will do a second PR related to changing the Earth radius. That one is not bit-for-bit so testing will be trickier, and we should discuss how to do it.

Please let me know what the schedule would be for testing and merging this. I understand that there are also other high priority items at the moment, but the EC30to60km meshes are needed as soon as possible so they can be tested and approved for the v2 release at the end of this month.

For EC60to30, here is the initial state just before the coastal merge and after #577 and #680. There are minor differences, but it looks reasonable.

Thanks for your work on this @xylar!

Thanks very much @sbrus89 and @mark-petersen for the review and for your testing.