Open h-vetinari opened 4 months ago
CC @chrisburr @vepadulano
IIRC the original problem is ROOT can't be cross-compiled, so @chrisburr was manually building ARM packages locally. I would assume we might have native support in conda-forge's CI now?
Thanks for the context. Should we re-attempt cross-compilation? I don't remember many cases where that wasn't ultimately solvable (even if we need to compile some extra stuff for the build agent architecture during building).
I would assume we might have native support in conda-forge's CI now?
No we don't yet. The macos-14
images are still running on amd64
agents at last check
IIRC the original problem is ROOT can't be cross-compiled, so @chrisburr was manually building ARM packages locally
Is there someplace where there's a summary of the issues on why ROOT can't be cross-compiled?
@chrisburr, were you doing something beyond just running python build-locally.py
? If not, then some of us who have Apple silicon machines could help out with this.
Can someone (yep, that might include myself, if instructed properly :wink: ) with an Apple Silicon be of any help to try to solve this or is there a fundamental technical problem preventing it ? (I have no idea how conda builds work, sorry, I'm just an end user currently)
(I'd love to be able to work again with a Conda Root package on my M1 Mac...)
If osx-arm64 is a supported platform (and going forward, that basically asks whether osx is a supported platform, as arm architecture takes over), then the builds should be green.
There have been many merges here with broken CI for osx-arm64, which leads to resolution errors downstream (e.g. https://github.com/conda-forge/stir-feedstock/pull/116).
IMO osx-arm64 needs to be either fixed (preferable) or dropped.