WIP: Build with flang 19

[h-vetinari]

Rebased #111 on top of #118, using the updated flang activation setup.

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe/meta.yaml) and found it was in an excellent condition.

[h-vetinari]

OK, I messed something up with the flang activation; I'll try again tomorrow with a fresh build (the build would pass if we set the flags manually, but I want to get to a point where that's not necessary).

[h-vetinari]

OK @Krande @traversaro, this now works as intended. It also works for flang 18, and curiously it doesn't even need llvm-openmp then (while for LLVM 19, it fails to find omp_lib.mod without it). I don't know the reasons for this, perhaps flang 18 just ignored certain openmp-related paths as unsupported?

There's also the fact that the environment then contains both llvm-openmp & intel-openmp (through mkl), which is not an amazing thing, but seems to work. IIRC, once we'll have the pytorch builds on windows, we should be able to get rid of intel-openmp across conda-forge.

[Krande]

@h-vetinari That's great news! What do you suppose is the remaining steps before we can update the global pinning feedstock for flang? I guess its this conda-forge/conda-forge-pinning-feedstock#1359 ?

Btw @traversaro in connection with this #112 I have managed to build mumps against scotch and metis for windows here. It would require changes to scotch and metis. I started a draft PR for scotch in conda-forge/scotch-feedstock#82 where we agreed we'd wait for the conda-forge pinning feedstock. So once the conda-forge pinning is updated I'll start making the necessary PR's

[traversaro]

Gret work @h-vetinari and @Krande !

[h-vetinari]

What do you suppose is the remaining steps before we can update the global pinning feedstock for flang? I guess its this conda-forge/conda-forge-pinning-feedstock#1359?

Yes, that's the right PR. I'm guessing building a couple key feedstocks and showing that flang 19 works on windows is going to go a long way in smoothing that path, so thanks for giving this a shot on a couple feedstocks already! I've successfully tested OpenBLAS & SciPy, but if you have more in mind then please go ahead and open PRs! :)

The remaining steps then are: wait for flang 19 GA (September), release it, and then start the migration (which will certainly need a bit of shepherding..., not least because regro/cf-scripts#1655 hasn't been tested in the wild yet).