Update to spack-stack 2.2

[mranst]

Description

Does not work, but there may be an obvious solution I'm missing. Builds successfully, but there are segmentation faults upon running in swell and ctests.

To try this, first:
source /discover/nobackup/manstett/spackstack_2.2_intel, then install jedi_bundle.

The following changes need to be made in the CMakeLists.txt for soca and fv3-jedi:

In fv3-jedi, the fms section needs to be:

# fms
find_package(FMS 2023.04 REQUIRED COMPONENTS R4 R8)
if (FV3_PRECISION MATCHES DOUBLE OR NOT FV3_PRECISION)
  add_library(fms ALIAS FMS::fms_r8)
  find_package(PkgConfig REQUIRED)
  pkg_check_modules(YAML REQUIRED IMPORTED_TARGET "yaml-0.1")
  target_link_libraries(FMS::fms_r8 INTERFACE PkgConfig::YAML)
else()
  add_library(fms ALIAS FMS::fms_r4)
  find_package(PkgConfig REQUIRED)
  pkg_check_modules(YAML REQUIRED IMPORTED_TARGET "yaml-0.1")
  target_link_libraries(FMS::fms_r4 INTERFACE PkgConfig::YAML)
endif()

In soca, near line 34 you need:

find_package( fms    2023.3.0 REQUIRED COMPONENTS R4 R8)
add_library(  fms ALIAS FMS::fms_r8)
find_package(PkgConfig REQUIRED)
pkg_check_modules(YAML REQUIRED IMPORTED_TARGET "yaml-0.1")

[Dooruk]

Hmm, is the FMS version correct? does spack 2.2 have a different one @mathomp4?

[mranst]

Yeah, from 1.9 it goes fms/2024.02 -> fms/2024.03-geos-constants

[mranst]

There's also fms/2025.03-gfs-constants and fms/2026.01.01-geos-constants

[Dooruk]

There's also fms/2025.03-gfs-constants and fms/2026.01.01-geos-constants

I see. I doubt that is relevant but just in case we need to update fms in soca or something.

Ok, i will give it a try next week. I think this is a Matt level problem though with spack-stack is concerned 😞

[Dooruk]

Tagging @metdyn @mer-a-o and @jeromebarre in case they have time to play around with these and GEOS-ESM/swell#679

[mathomp4]

Yeah, from 1.9 it goes fms/2024.02 -> fms/2024.03-geos-constants

You want to use fms@2024.03 for anything GEOS related. I do have a series of PRs (see GEOS-ESM/GEOSgcm#933) to move GEOS to use fms@2026.01.01 but I've been waiting for GEOS to "settle down" as this is a non-zero-diff change. I'd prefer to have a tag of GEOS with FMS 2024.03 and then an "as identical as possible" version with FMS 2026.

Note: with FMS 2026, we go from the two FMS::fms_r8 and FMS::fms_r4 targets to one single mixed-precision FMS::fms target. So that changes quite a bit of CMake.