Installation instructions for Extended MODFLOW don't work

[geo-admin-joe]

I'm trying to follow the installation instructions from the Extended MODFLOW User Guide. What I did:

1. git clone https://github.com/MODFLOW-USGS/modflow6.git
2. Created mf6xtd_environment.yml as decribed.
3. conda env create -f mf6xtd_environment.yml
4. conda activate mf6xtd
5. cd modflow6
6. meson setup builddir -Ddebug=false -Dextended=true --prefix=$(pwd) --libdir=bin

So far everything worked without error.

Now I call: meson install -C builddir

Output:

(mf6xtd) root@PC12365:~/modflow6# meson install -C builddir
ninja: Entering directory `/root/modflow6/builddir'
[1/140] Compiling Fortran object src/libmf6core.a.p/Model_GroundWaterFlow_gwf-api.f90.o
f951: Warning: Nonexistent include directory '-fdiagnostics-color=always' [-Wmissing-include-dirs]
[2/140] Compiling Fortran object src/libmf6core.a.p/Model_GroundWaterEnergy_gwe-esl.f90.o
f951: Warning: Nonexistent include directory '-fdiagnostics-color=always' [-Wmissing-include-dirs]

... more "Nonexistent include directory '-fdiagnostics-color=always'" warnings ... 

[49/140] Compiling Fortran object src/libmf6core.a.p/Solution_PETSc_PetscConvergence.F90.o
FAILED: src/libmf6core.a.p/Solution_PETSc_PetscConvergence.F90.o src/libmf6core.a.p/petscconvergencemodule.mod
gfortran -Isrc/libmf6_external.a.p -Isrc/libmf6core.a.p -Isrc -I../src -I/root/.conda/envs/mf6xtd/include -I -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -std=f2018 -O3 -fall-intrinsics -pedantic -cpp -Wcharacter-truncation -Wno-unused-dummy-argument -Wno-intrinsic-shadow -Wno-maybe-uninitialized -Wno-uninitialized -D__linux__ -ffpe-summary=overflow -ffpe-trap=overflow,zero -D__WITH_PETSC__ -D__WITH_MPI__ -D__WITH_NETCDF__ -fPIC -pthread -Jsrc/libmf6core.a.p -o src/libmf6core.a.p/Solution_PETSc_PetscConvergence.F90.o -c ../src/Solution/PETSc/PetscConvergence.F90
f951: Warning: Nonexistent include directory '-fdiagnostics-color=always' [-Wmissing-include-dirs]
../src/Solution/PETSc/PetscConvergence.F90:48:32:

   48 |       KSPConvergedReason :: flag
      |                                1
Error: Derived type 'ekspconvergedreason' at (1) is being used before it is defined

... more "Nonexistent include directory '-fdiagnostics-color=always'" warnings ... 

[61/140] Compiling Fortran object src/libmf6core.a.p/Model_GroundWaterFlow_gwf-lak.f90.o
f951: Warning: Nonexistent include directory '-fdiagnostics-color=always' [-Wmissing-include-dirs]
ninja: build stopped: subcommand failed.
Could not rebuild /root/modflow6/builddir

What should I do?

[wpbonelli]

Hi @geo-admin-joe what version of meson do you have? I'd recommend using the pixi environment if you can since it has locked dependencies. But try meson==1.7.1 if you have something different.

Thanks for the heads up about the link, I fixed it in the wiki. Here is the correct one

[geo-admin-joe]

Thank you for the quick reply! I tried meson==1.7.1, but it made no difference.

[mjr-deltares]

Hi @geo-admin-joe , there have been breaking changes in a recent PETSc release. Can you try pinning the PETSc version in the yml:

- petsc=3.22

[geo-admin-joe]

OK, nice, I'm one step ahead. The error Derived type 'ekspconvergedreason' at (1) is being used before it is defined is gone now. 👍 Thank you!

The hundreds of f951: Warning: Nonexistent include directory '-fdiagnostics-color=always' warnings are still there, though.

[mjr-deltares]

I remember coming across that warning as well. I believe it's an issue with the (PETSc or MPI, not sure) package config file (*.pc) that comes with the installation. I will have a look into it later this week and get back to you.

[mjr-deltares]

Hi @geo-admin-joe ,

The problem is with the package config file for netcdf-fortran:

<path to envs>\envs\mf6xtd\lib\pkgconfig\netcdf-fortran.pc

The Fortran module include directory is not specified. In the pc file you will see:

fmoddir=

...

Cflags: -I${includedir} -I${fmoddir}

Just delete the line with fmoddir= and remove the 2nd Cflag -I${fmoddir} from the definition. For me that takes care of all the warnings.

[geo-admin-joe]

This did the trick. Thank you so much!

Just a quick note if someone else with the same problem reads this: The changes to netcdf-fortran.pc still need to be exported/applied. The quickest way for me was a simple reboot.

[wpbonelli]

@geo-admin-joe do you by chance have netcdf-fortran 4.6.1? @mjr-deltares and I found that the netcdf-fortran.pc file generated for 4.6.2 looked ok.

[geo-admin-joe]

Yes, indeed. I tried to use 4.6.2, but conda told me it's incompatible with petsc 3.22. I assume that's why netcdf-fortran was installed in version 4.6.1 automatically.