@ShunLiu-NOAA , @hu5970 , and @ManuelPondeca-NOAA : Are we OK with setting the default MGBF build option to OFF?

@RussTreadon-NOAA it is OK for all RRFS test now.

@RussTreadon-NOAA Turning off the MGBF option , currently, would cause the building of GSI fail, because, unfortunately, the mgbf codes in gsi are not coordinated with the compiler option as the GSDCLOUD subroutines in src/gsi (like gsdcloudanalysis.F90 the "ifdef RR_CLOUNDANALYSIS" block ) does. So, A PR for this issue would need to include adding similar treatments (but more efforts needed for the smaller pieces of mgbf codes in hybrid_ensemble_isotropic.F90) for mgbf codes in gsi.

Thank you @TingLei-NOAA for explaining that setting BUILD_MGBF to OFF would cause the GSI build to fail. I did not realize this. This is unfortunate.

We should not have allowed PR #700 to be merged into develop as is MGBF's addition to the GSI should have followed the approach used by the GSD cloud analysis. GSI realizations which need the GSD cloud analysis set BUILD_GSDCLOUD on. The default is off. The same should be true for MGBF.

GFS v16 does not and GFS v17 will not use MGBF. However, we must build MGBF as part of the global GSI.

@RussTreadon-NOAA Your points are completely understood.
If you and other colleagues agree, I could make a draft pr to do the similar treatment to mgbf codes in gsi to allow the mgbf option to be turned off as GSDCLOUD. My concern for such a treatment for mgbf is those mgbf codes are scattered and might cause the treatments using directives looks not clean.
Please let me know if we should explore this change.

I don't know the MGBF code so I don't know how extensive the changes may be. Thus, a draft PR would be useful. That said, do not read this comment as a request to prepare a draft PR. I am not tasking anyone to do anything. Please discuss among staff the appropriate action(s) to take.

@RussTreadon-NOAA Thanks for your inputs! I will keep you updated of what is going on with this topic!

@RussTreadon-NOAA @ShunLiu-NOAA , we have a draft PR to show the changes done for MGBF to be able to be turned off. Please let us know if that changes in that draft PR are regarded as the solution for this PR and could proceed to the PR process, or a better solution could be sought.

Thank you @TingLei-NOAA for making these changes. The changes look straightforward. Do the changes yield the desired behavior

1. successfully build and run ctests without MGBF (default)
2. successfully build and run with MGBF when USE_MGBF=ON

1 is easily tested. 2 requires you to run a test which exercises MGBF.

@RussTreadon-NOAA The current regression tests are being run with MGBF turned off.
A MGBF test as discussed in another thread is being discussed and being prepared but might need some time .

@RussTreadon-NOAA The current regression tests are being run with MGBF turned off. A MGBF test as discussed in another thread is being discussed and being prepared but might need some time .

Yes, it is important to add a MGBF ctest if we intend to use MGBF via an operational realization of the GSI. Until this test is available, you should run an offline test to confirm that 2 is valid.

@TingLei-NOAA , what is the status of this issue?

@RussTreadon-NOAA I had put the corresponding PR to be on "draft", while ,as I reported , using the current compiler on wcoss2,,that PR could fail the ctests even when the option is on. I will request an extension to keep this issue open and I will spend more time to work on it.

@TingLei-NOAA , thanks for the update.