gfs_utils build fails on Orion

[RussTreadon-NOAA]

Expected behavior
The gfs_utils build should succeed when executing sorc/build_all.sh on Orion.

Current behavior
The gfs_utils build fails on Orion with the following message:

 .... Building gfs_utils ....
Fatal error in building gfs_utils.
The log file is in /work/noaa/da/Russ.Treadon/git/global_workflow/test/sorc/logs                              /build_gfs_utils.log
FATAL BUILD ERROR: Please check the log file for detail, ABORT!

The following error message is found in sorc/logs/build_gfs_utils.log

CMake Error at CMakeLists.txt:44 (find_package):
  By not providing "Findg2.cmake" in CMAKE_MODULE_PATH this project has asked
  CMake to find a package configuration file provided by "g2", but CMake did
  not find one.

  Could not find a package configuration file provided by "g2" with any of
  the following names:

    g2Config.cmake
    g2-config.cmake

  Add the installation prefix of "g2" to CMAKE_PREFIX_PATH or set "g2_DIR" to
  a directory containing one of the above files.  If "g2" provides a separate
  development package or SDK, be sure it has been installed.

Machines affected
Orion

To Reproduce

1. git clone https://github.com/NOAA-EMC/global-workflow.git .
2. cd sorc
3. ./checkout.sh
4. ./build_all.sh. If one wants to more quickly see the failure, edit gfs_build.cfg so that all builds are off except gfs_utils

Possible Implementation
According to build_gfs_utils.log the g2 package is not found. gfsutils_common.lua loads, by default, g2/3.4.5 However, logs/build_gfs_utils.log does not include g2 in the module list. The module use and load commands in gfs_utils.fd/ush/build.sh were executed in an interactive Orion session. It was found that g2/3.4.5 requires hpc-intel/2022.1.2. This is problematic. gfsutils_orion loads hpc-impi/2018.4.

Correction of the problem reported in this issue will require changes be made in gfs-utils.

[KateFriedman-NOAA]

I just hit this too and have a remedy. Will fix in gfs-utils. Thanks!

[KateFriedman-NOAA]

FYI: NOAA-EMC/gfs-utils#14

[KateFriedman-NOAA]

FYI: NOAA-EMC/gfs-utils#15

[KateFriedman-NOAA]

New hash of gfs-utils (56e6880) resolves the build error on Orion. That hash also updated gfs-utils to use module reset instead of module purge on most of the machines. @WalterKolczynski-NOAA will have a companion PR to make a similar module reset update to global-workflow build scripts and move to this new hash shortly. @RussTreadon-NOAA you can checkout that new hash on Orion for now to get it built in your clone.

[RussTreadon-NOAA]

@KateFriedman-NOAA , I updated my working copy of gfs_utils.fd to 56e6880 and reran g-w build_all.sh for gfs_utils alone. The build was successful.

Orion-login-2:/work/noaa/da/Russ.Treadon/git/global_workflow/test/sorc$ ./build_all.sh
 .... Skip building gsi and enkf ....
 .... Skip building gsi utilities ....
 .... Skip building GDASApp  ....
 .... Skip building gsi monitor ....
 .... Skip building gldas ....
 .... Building gfs_utils ....

 .... Build system finished ....

I did not attempt to execute any gfs_util apps or scripts.

[KateFriedman-NOAA]

Cool, thanks for confirming @RussTreadon-NOAA !