Thoughts on adding pfunit-based unit tests to CMEPS?

[billsacks]

I'm planning to add some code to CMEPS soon for water isotopes and water tracers more generally. I may want unit tests for some of this code - and the addition of unit tests would be in line with one of the deliverables of the CSSI funding for this work.

Would people find it acceptable to start adding pFUnit-based unit tests of the Fortran code here?

One thing that will differ from other code bases where I have added these pFUnit-based unit tests in the past is that CMEPS already has a CMake-based build system. In other projects, like CTSM, the CMakeLists.txt files are used exclusively for the unit test build. My initial thinking is that it will probably be okay to share the CMakeLists.txt files between the "real" build and the unit test build, though it may require some constraints / logic:

• I would probably set up the unit test build to use the full CMEPS library, rather than building just the source files needed for a given unit test (so most of the build can be shared between the real build and the unit test build, rather than having different builds for each of those cases)
• I would add conditional logic in the top-level CMakeLists.txt file that adds test-specific subdirectories and other pFUnit-specific things based on whether a test flag is given to the cmake build.

I'm not sure if I'll necessarily do this as part of the upcoming work (partly because I don't want to go down any deep rabbit holes), but want to start this discussion and hopefully get some consensus before I spend too long working on anything.

(@jedwards4b @DeniseWorthen @mvertens @uturuncoglu )

[jedwards4b]

I would be okay with this as long as the pfUnit tests are included in the github workflow.

[mvertens]

@billsacks - I am fine with this.

[DeniseWorthen]

@billsacks I read through the information on the pFunit github site. The only concern I have is adding a dependency which would be unavailable on the operational platforms. I don't know if that could be an issue w/ pFunit (it appears to be an officially supported repo), but I imagine in any case pFUnit would not break CMEPS if it were not available---and UFS does have it's own cmake build for CMEPS.

[billsacks]

Thanks for your thoughts, @jedwards4b , @mvertens and @DeniseWorthen .

@jedwards4b - yes, great idea to add any pFUnit tests to the GitHub workflow.

@DeniseWorthen - yes, as you say, I imagine setting this up so that pFUnit is optional, and would only be looked for when building/running unit tests, not when building CMEPS for production use. Though also, yes, pFUnit is supported by Tom Clune and others at NASA-Goddard.