While the original code changes in https://github.com/NOAA-EMC/NEMS/pull/7
are valid and standard Fortran and work on macOS with gcc-8.3.0 and gcc-9.1.0, they lead to a model crash on Cheyenne with gcc-8.3.0:
terminate called after throwing an instance of 'std::length_error' what(): basic_string::_M_create
The solution is to revert to the failsafe version that we have used successfully in the past for both Intel and GNU.
I tested those changes to work for both Intel and GNU on Cheyenne and I would advocate for merging them as soon as we can (possibly before kicking off the final round of tests for https://vlab.ncep.noaa.gov/code-review/#/c/19568).
This was a red herring, I am closing this PR. The proposed change doesn't fix the problem, it rather seems to be related to the ESMF library used on Cheyenne/GNU (8.0.0bs40).
This PR fixes an issue with a code change introduced in https://github.com/NOAA-EMC/NEMS/pull/7 (src/module_MEDIATOR.F90).
While the original code changes in https://github.com/NOAA-EMC/NEMS/pull/7 are valid and standard Fortran and work on macOS with gcc-8.3.0 and gcc-9.1.0, they lead to a model crash on Cheyenne with gcc-8.3.0:
The solution is to revert to the failsafe version that we have used successfully in the past for both Intel and GNU.
I tested those changes to work for both Intel and GNU on Cheyenne and I would advocate for merging them as soon as we can (possibly before kicking off the final round of tests for https://vlab.ncep.noaa.gov/code-review/#/c/19568).