Error with `SMS_D_Lh4.ne4pg2_ne4pg2.F2010-SCREAMv1.pm-cpu_gnu.eamxx-prod` Fortran runtime error: Pointer argument 'cospout' is not associated

[ndkeen]

With following GNU DEBUG test:

SMS_D_Lh4.ne4pg2_ne4pg2.F2010-SCREAMv1.pm-cpu_gnu.eamxx-prod

59: At line 502 of file /global/cfs/cdirs/e3sm/ndk/repos/ndk_mf_alvarez-module-updates-feb2025/components/eam/src/physics/cosp2/external/src/cosp.F90
59: Fortran runtime error: Pointer argument 'cospout' is not associated

The non-debug test completes.

Also, Intel compiler DEBUG completes.

[mahf708]

xref: https://github.com/E3SM-Project/E3SM/issues/6951

[ndkeen]

Noting that after NERSC maintenance, and looking at the new issues with DEBUG cases, this case still fails in the same way.

[ndkeen]

Same error with June18th checkout

[ndkeen]

I think we can correct this error with:

    if (associated(cospOUT%isccp_meantb)) then
        if (size(cospOUT%isccp_meantb) .eq. stop_idx) lrttov_cleanUp = .true.
    end if

It fixes this issue (will make PR), but then I think there may be other similar issues in other parts of the code that I could make another PR for. Ack, I now see this code is in external submodule so I can't easily make a PR.

Ben notes that with eam, we use components/eam/src/physics/cosp2/local/cosp.F90, but eamxx uses components/eam/src/physics/cosp2/external/src/cosp.F90.

[bartgol]

Yeah, the fix seems simple. But the submodule org makes this a longer process. I'm ok switching to a local copy like EAM. Does it make sense to still have a cosp submodule then? Who uses it besides eamxx, given that EAM has its local copy?

[brhillman]

Sorry I didn't look at this sooner, but yes we would still need the cosp submodule because this is only a very small piece of COSP, all the actual simulators are still used from the submodule, this is just interface code.

[brhillman]

Also, it looks like this is fixed upstream in COSP, so we probably just need a PR to pull in the latest COSP dev. See https://github.com/CFMIP/COSPv2.0/blob/master/src/cosp.F90:

    if (associated(cospOUT%isccp_meantb) .and. size(cospOUT%isccp_meantb) .eq. stop_idx) lrttov_cleanUp = .true.