Modifications to run workflow with unmodified executable names from each repository

[mkavulich]

Note: This PR must be merged at the same time as SRW App PR#27

DESCRIPTION OF CHANGES:

The new top-level cmake build for the SRW App (SRW App PR#27) results in some executables having different names. This PR makes modifications that

1. Allow the workflow to run successfully with the new cmake build and its different executable names, and
2. Allow back-compatibility with the old build system to allow for a gradual transition to new build system

This PR also explicitly disallows running the workflow without CCPP, which we decided against supporting several months ago. I don't think the capability even works so this shouldn't effect anyone at this time.

TESTS CONDUCTED:

• Cheyenne: Build and end-to-end test ("DOT_OR_USCORE" test case) was successful on Cheyenne with intel, both for the cmake build and the old build script (that will soon be deprecated). Path to tests: /glade/scratch/kavulich/UFS_CAM/testing/SRW_PR_27/expt_dirs/
• Hera: Build and end-to-end tests successful (aside from expected failures). Path to tests: /scratch2/BMC/det/kavulich/workdir/SRW_PR_27/expt_dirs
  • Currently debugging end-to-end tests on Hera; there is some kind of library problem related to the latest ufs-weather-model commit (ufs-community/ufs-weather-model@f68dcdb) Fixed
• Jet: Build test was successful. Machine is having disk errors that are preventing me from running the end-to-end tests.

ISSUE:

It was not the primary aim of this PR, but it does partially resolve #196

[mkavulich]

A question about backwards compatibility: in its current form, this PR would mean that the old method of building the workflow will no longer work: only the cmake top-level build will create the correct executable names. It would be a quick fix to the "install_all.sh" script to allow backward compatibility, is this something we want implemented?

[jwolff-ncar]

I personally don't think we need to support it. But if others feel we should, perhaps it could be only in develop (for a period of time) and not the release branch?

…

On Wed, Oct 7, 2020 at 2:22 AM Michael Kavulich ***@***.***> wrote: A question about backwards compatibility: in its current form, this PR would mean that the old method of building the workflow will no longer work: only the cmake top-level build will create the correct executable names. It would be a quick fix to the "install_all.sh" script to allow backward compatibility, is this something we want implemented? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <https://github.com/NOAA-EMC/regional_workflow/pull/304#issuecomment-704776468>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFN5ZDCK6XF7YL6ZKS56FDLSJQQK3ANCNFSM4SHBQMKQ> .

-- Jamie K. Wolff National Center for Atmospheric Research (NCAR) Research Applications Laboratory (RAL) Phone: 303.497.2812 Email: jwolff@ucar.edu *My working day may not be your working day. Please do not feel obliged to reply to this email outside of your normal working hours.*

[mkavulich]

The library problem referenced in my PR message has been resolved (by a change in the other PR in ufs-srweather-app). Build and end-to-end test ("DOT_OR_USCORE" test case) was successful on Cheyenne with intel, both for the cmake build and the old build script (that will soon be deprecated). Hera build and end-to-end suite of tests are also successful now. Jet is awaiting some final testing that is being delayed due to some sort of disk issue; if I can't get these tests run by tomorrow afternoon I would propose merging anyway and fixing any problems afterwards....this code slush has dragged on too long!

[JulieSchramm]

Looks good.

[gsketefian]

Would be good to change the path
${SR_WX_APP_TOP_DIR}/src/ufs_weather_model
in this if-statement to ${UFS_WTHR_MDL_DIR} so there's less hard-coding.
Same thing on line 609 below.

[gsketefian]

Even better, to reduce repetition/hard-coding, I think this will work:

exec_fp="${SR_WX_APP_TOP_DIR}/bin/${exec_fn}"
#Check for the old build location for fv3 executable
if [ ! -f "${exec_fp}" ]; then
  exec_fp_alt="${UFS_WTHR_MDL_DIR}/build/${exec_fn}"
  if [ ! -f ${exec_fp_alt}" ]; then
    print_err_msg_exit "\
The executable (exec_fp) for running the forecast model does not exist:
  exec_fp = \"${exec_fp}\"
Please ensure that you've built this executable."
  else
    exec_fp="${exec_fp_alt}"
  fi
fi

[mkavulich]

Good suggestion, made that edit. Running one final end-to-end test and then will merge both PRs.