Fix CUDA awareness

[leofang]

Checklist

• [x] Used a personal fork of the feedstock to propose changes
• [x] Bumped the build number (if the version is unchanged)
• [ ] Reset the build number to 0 (if the version changed)
• [x] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• [x] Ensured the license file is being packaged.

Close #83. Fix a silly mistake that I made in #80...😞

[conda-forge-linter]

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) and found it was in an excellent condition.

[leofang]

@conda-forge-admin, please rerender

[jakirkham]

Should the run_constrained in meta.yaml also be updated?

[github-actions[bot]]

Hi! This is the friendly automated conda-forge-webservice. I tried to rerender for you, but it looks like there was nothing to do.

[leofang]

Should the run_constrained in meta.yaml also be updated?

No, because we can support any CUDA version. Open MPI does not care. I set it to 9.2 only because it's the oldest version available on CF.

[jakirkham]

Ok didn't know if we would run into issues since we are actually building against CUDA 10.2

[leofang]

Right, we made sure of that earlier. Open MPI does not rely on any CUDA runtime API, only driver ones. It dlopen libcuda.so at runtime. It has been the case since CUDA 6 (?). There are some weird CUDA version check at build time that should have been removed (https://github.com/open-mpi/ompi/issues/7334).

[leofang]

Travis is dead 😞

@conda-forge-admin please restart ci

[leofang]

osx segfaults...?

[leofang]

@conda-forge-admin please restart ci

[github-actions[bot]]

Hi! This is the friendly conda-forge automerge bot!

I considered the following status checks when analyzing this PR:

• linter: passed
• drone: passed
• travis: failed
• azure: passed

Thus the PR was not passing and not merged.

[leofang]

I manually restarted Travis, but if it fails the 3rd time, I will merge anyway. It's been in a poor state in weeks and not getting any better: https://gitter.im/conda-forge-ppc64le/Lobby

[leofang]

...and it failed again as I typed.

[leofang]

Merging, and will restart manually until the package is out, although the ppc64le package built on Travis is not impacted by my silly mistake, so it's fine to not upload a new one. In the near future let's look for other alternatives, maybe cross-compiling?

[leofang]

Merging, and will restart manually until the package is out

The merge commit built successfully.

[jakirkham]

Sorry for missing this before, but should these lines have been updated to CUDA 10.2 as well?

https://github.com/conda-forge/openmpi-feedstock/blob/fde85a81ef3e94a6714d4227e9968e59da19300e/recipe/meta.yaml#L47-L48