Fix memory errors: uninitialized variables, and NetCDF writes to arrays that are too small

[SamuelTrahanNOAA]

Description

This fixes two memory errors detected with valgrind, and one missing error check (errmsg, errflg). See the "issues addressed" for details.

Issue(s) addressed

• fv3atm: fixes Uninitialized Grid variables in GFS_typedefs.F90 grid_create #667
• ccpp-physics: 32-bit physics crashes in aerinterp due to type mismatch in NetCDF calls. ufs-community/ccpp-physics#83
• ccpp-physics: aerinterp routines do not pass NetCDF errors up through errmsg and errflg ufs-community/ccpp-physics#84

Testing

How were these changes tested?

A 32-bit physics version of the RRFS conus13km regression tests, using valgrind.

What compilers / HPCs was it tested with?

Hera gnu and intel

Are the changes covered by regression tests? (If not, why? Do new tests need to be added?)

There are new regression tests

Have the ufs-weather-model regression test been run? On what platform?

Yes. Hera.

Will the code updates change regression test baseline? If yes, why? Please show the baseline directory below.

No, but there are new tests.

Please commit the regression test log files in your ufs-weather-model branch

Will do.

Dependencies

• waiting on ccpp-physics Fix bug in aerinterp read routines that break 32-bit physics and pass NetCDF errors up through errflg&errmsg ufs-community/ccpp-physics#85

[junwang-noaa]

@SamuelTrahanNOAA It looks to me the error-check in ufs-community/ccpp-physics#83 is to prevent the error of read(4) array from reading real(8) data. I am wondering what the solution is, so you will need a new netcdf file with real(4) data in order to run the 32bit physics?

[SamuelTrahanNOAA]

@SamuelTrahanNOAA It looks to me the error-check in ufs-community/ccpp-physics#83 is to prevent the error of read(4) array from reading real(8) data. I am wondering what the solution is, so you will need a new netcdf file with real(4) data in order to run the 32bit physics?

My fix for the type conversion error was to use the NetCDF Fortran 90 interface. The Fortran 90 interface does automatic type conversion. Previously, that code used the Fortran 77 NetCDF interface.

[junwang-noaa]

Got it, thanks

[SamuelTrahanNOAA]

Quoting the documentation of NF90_GET_VAR,

Data values are converted from the external type of the variable, if necessary.

They define "external type of a variable" as the type it has in the file. An exception to this rule is string arrays, which have no automatic type conversion.

[SamuelTrahanNOAA]

The ccpp-physics PR has been merged. I have update the submodule pointer and reverted the .gitmodules. We're waiting for this PR to be merged so we can finish ufs-community/ufs-weather-model#1797

[zach1221]

@jkbk2004 can you merge this PR, please? I'm not authorized in this case.