Compilation warnings in non-debug mode for GFSv17

[CatherineThomas-NOAA]

NCO standards say “Did code compile without any error or warning messages, i.e., is it verified that no compiler options were used to suppress any error or warning messages?"

When we compile GDASApp in normal mode (i.e. not debug mode), we still have many warning messages that fall into several types:

GDASApp source code (saber, fv3jedi, ioda)

• 5 cases in saber: warning #47: incompatible redefinition of macro "ERR"
• 3 cases in fv3jedi: warning #6178: The return value of this FUNCTION has not been defined. (other cases in tmp/ifort*)
• 13 cases in fv3jedi: warning #6843: A dummy argument with an explicit INTENT(OUT) declaration is not given an explicit value.
• 3 cases in ioda: warning #1478: function "gsl::span<T>::span(gsl::span<T>::reference) [with T=const std::__cxx11::string]" (declared at line 3694 of "/apps/prod/gsl-lite/v0.40.0/include/gsl/gsl-lite.hpp") was declared deprecated
• 56 cases in ioda, 1 case in ufo: warning #5462: Global name too long, shortened from: ioda_variable_creation_parameters_mod_mp_IODA_VARIABLE_CREATION_PARAMETERS_SET_FILL_VALUE_INT64 to: _variable_creation_parameters_mod_mp_IODA_VARIABLE_CREATION_PARAMETERS_SET_FILL_VALUE_INT64
• 677 cases in ioda, saber, oops: warning #858: type qualifier on return type is meaningless
• 7 cases in saber: warning #823: reference is to variable "jl" (declared at line 1042) -- under old for-init scoping rules it would have been variable "jl" (declared at line 1074)

Other warnings for thing that might be out of our control? Cmake/ifort/fms/eckit

• 26 cases: CMake Warning: Policy CMP0144 is not set: find_package uses upper-case <PACKAGENAME>_ROOT variables.
• 8 cases: warning: relocation in read-only section .trace
• 8 cases: warning: creating DT_TEXTREL in a shared object
• 5 cases: fms_platform.h(88): #warning: macro redefined: QUAD_KIND (fms)
• 200 cases: warning #2651: attribute does not apply to any entity (eckit)
• 503 cases: “ifort: command line warning #10121: overriding '-qopenmp-stubs' with '-qopenmp'”

This build was compiled using the new atlas and can be found here on Dogwood: /lfs/h2/emc/global/noscrub/travis.j.elless/global_workflow/v17_atlas_test2/global-workflow

A saved version of the build output log can be found here on Dogwood: /lfs/h2/emc/da/noscrub/catherine.thomas/tmp/build_gdas.log

I’d appreciate any help digging through these, seeing what warnings are reasonable to take care of and which ones aren’t.

cc @CoryMartin-NOAA @guillaumevernieres @shlyaeva @danholdaway @DavidNew-NOAA @RussTreadon-NOAA

[guillaumevernieres]

@CatherineThomas-NOAA I'm planning to spend some time on this today.

[guillaumevernieres]

In case somebody feels like tackling some of these warnings, here's what I'm working on (branches pushed to jcsda-internal):

• saber: hotfix/warnings-c8e0e3c3.
• gdas-utils: hotfix/warnings-508c40a4 (WIP). The changes are significant, it needs testing and if we don't see these warnings in WCOSS2, let's not include them.
• ufo: hotfix/warnings-bb78f475
• fv3-jedi: hotfix/warnings-777c784b
• working on ioda now, that should keep me busy for a while. Warnings in the 100's ... Oops

[RussTreadon-NOAA]

@CatherineThomas-NOAA and @TravisElless-NOAA

Dogwood /lfs/h2/emc/global/noscrub/travis.j.elless/global_workflow/v17_atlas_test2/global-workflow no longer exists.

If we want to replicate the above list of compiler warnings, I assume we start from a clone of g-w dev/gfs.v17. The current head of dev/gfs.v17 points at gdas.cd @ 508c40a. The build.sh in this hash does not include a DEBUG build option. How was the GDASApp built in DEBUG mode? Was GDASApp build.sh locally modified?

[RussTreadon-NOAA]

Looks like I can use variant of NOAA-EMC/global-workflow#5026 for the GDASApp debug build.

[TravisElless-NOAA]

@RussTreadon-NOAA sorry about that. We got a new library installed and didn't want to confuse myself in testing and forgot I sent that directory as a log guide. I have rebuilt things on dogwood at /lfs/h2/emc/global/noscrub/travis.j.elless/global_workflow/v17_atlas_test/gfs.v17.0.0/. Changes I had made locally in gdas.cd for the build, I have captured in https://github.com/TravisElless-NOAA/GDASApp/tree/feature/v17_ops_modules. I was waiting on the official python virtual environment to be installed before creating a PR as the global workflow marine jobs will not run with the currently loaded modules in this branch.

Relatedly I also think your NOAA-EMC/global-workflow#5026 (comment) will likely get resolved using the intel 19 compiler option from this updated module file instead of the default intel-oneapi version.

Sorry for all the moving parts we have going on here.

[RussTreadon-NOAA]

Thank you @TravisElless-NOAA for sharing these important details. I lost track of where we were with GFS v17 pre-implementation work. It's good to get reoriented.

I copied your Dogwood gfs.v17.0.0 directory to my space, replaces sorc/gdas.cd with release/gfs.v17, and copied the three modified files from feature/v17_ops_modules into my working copy. Changes from g-w PR #5026 were also added.

You are right. The debug build of this combination of changes works on Dogwood. Caveat: The use of "works" in the previous sentence only applies to compilation. I did not attempt to execute the resulting gdas.x.

[RussTreadon-NOAA]

@CatherineThomas-NOAA : Does EE2 specify the build for which we need to check compiler warnings? Are we looking for warnings in the Release build, the Debug build, or both?

[CatherineThomas-NOAA]

@RussTreadon-NOAA - Our interpretation of the rules is that:

1. We should be checking for the compiler warnings in the Release build, and
2. The Debug build needs to be able to at least compile and run successfully (warnings are less important here)

[RussTreadon-NOAA]

@CatherineThomas-NOAA . Got it. Thank you.

Both Release and Debug builds of Dogwood /lfs/h2/emc/da/noscrub/russ.treadon/git/global-workflow/gfs.v17.0.0/sorc/gdas.cd fail for me with the error

/lfs/h2/emc/da/noscrub/russ.treadon/git/global-workflow/gfs.v17.0.0/sorc/gdas.cd/bundle/soca/src/soca/../soca/AnalyticInit/AnalyticInit.h(13): catastrophic error: cann\
ot open source file "ufo/AnalyticInitBase.h"
  #include "ufo/AnalyticInitBase.h"
                                   ^

compilation aborted for /lfs/h2/emc/da/noscrub/russ.treadon/git/global-workflow/gfs.v17.0.0/sorc/gdas.cd/bundle/soca/src/soca/AnalyticInit/AnalyticInit.cc (code 4)
make[2]: *** [soca/src/soca/CMakeFiles/soca.dir/build.make:76: soca/src/soca/CMakeFiles/soca.dir/AnalyticInit/AnalyticInit.cc.o] Error 4
make[2]: Leaving directory '/lfs/h2/emc/da/noscrub/russ.treadon/git/global-workflow/gfs.v17.0.0/sorc/gdas.cd/build'
make[1]: *** [CMakeFiles/Makefile2:6632: soca/src/soca/CMakeFiles/soca.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....

@TravisElless-NOAA , is it premature to attempt building sorc/gdas.cd with your modified modulefiles/GDAS/wcoss2.intel.lua?

[TravisElless-NOAA]

@RussTreadon-NOAA I have gotten both develop and dev/gfs.v17 versions of GDASApp to build with that modified modulefiles/GDAS/wcoss2.intel.lua utilizing the global workflow build_all.sh gdas gsi gfs. I believe @CoryMartin-NOAA has had success as well previously using this updated module list. If memory serves me correct, I have run into this error before doing this and determined something did not clone properly for some reason. Doing a fresh clone again somehow got it to work for me, so that would be my first troubleshooting suggestion.

[RussTreadon-NOAA]

@RussTreadon-NOAA I have gotten both develop and dev/gfs.v17 versions of GDASApp to build with that modified modulefiles/GDAS/wcoss2.intel.lua utilizing the global workflow build_all.sh gdas gsi gfs. I believe @CoryMartin-NOAA has had success as well previously using this updated module list. If memory serves me correct, I have run into this error before doing this and determined something did not clone properly for some reason. Doing a fresh clone again somehow got it to work for me, so that would be my first troubleshooting suggestion.

@TravisElless-NOAA
Just to close this thread: A fresh clone g-w dev/gfs.v17 at ab5f38dc2 with updates from your GDASApp feature/v17_ops_modules and Cory's g-w feature/build_toolchain successfully builds in Release and Debug mode on Dogwood.