search

conda-forge / osqp-feedstock

A conda-smithy repository for osqp.
BSD 3-Clause "New" or "Revised" License
0 stars 13 forks source link

Enable ppc64le and update to 0.6.1 #14

Closed jayfurmanek closed 5 years ago

jayfurmanek commented 5 years ago

Checklist

Enable building for ppc64le arch using Travis CI

jayfurmanek commented 5 years ago

@conda-forge-admin, please rerender

conda-forge-linter commented 5 years ago

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.

jayfurmanek commented 5 years ago

@conda-forge-admin, please rerender

jayfurmanek commented 5 years ago

@conda-forge-admin, please rerender

h-vetinari commented 5 years ago

Hey @jayfurmanek Cool that you're enabling ppc64le builds!

I have to admit that it's an architecture that's new to me, and I wouldn't know how to debug it if problems were to arise. For example, we're currently struggling to debug the windows builds in #11 / #12, and I'm hesitant to add something to the CI that might get us similarly stuck.

As such, I'd ask you to try if you can get this branch (or better: a new PR) to work also with these two commits, and see if ppc64le would work there.

Furthermore, if you're willing to take care of ppc64le-builds, it would be reassuring to have you in the list of recipe maintainers (in this case, just add your GH-handle at the end of meta.yaml), so you'd hopefully help us get unstuck if ppc64le-builds were to fail in the future.

PS. What's the reason for adding make (beyond debugging the build) if there is one? Not sure how make/cmake behave differently on different platforms, but at least on non-ppc64le-systems we shouldn't need make. (i.e. add # [ppc64le] after make).

jayfurmanek commented 5 years ago

Hi @h-vetinari, Sure, I'll merge the 0.6.1 update here, although I'm not sure I can help with any Windows stuff, I can certainly can help maintain the Linux ppc64le side.

Make is required for Linux, at least. Cmake generates makefiles and make is needed to do the build. I think it *was? baked into the anvil image and maybe isn't anymore. I'll look into the full story there. Maybe a # [unix] is more appropriate.

h-vetinari commented 5 years ago

@jayfurmanek: Sure, I'll merge the 0.6.1 update here, although I'm not sure I can help with any Windows stuff, I can certainly can help maintain the Linux ppc64le side.

There's no expectation that you'll have to solve the windows problems, but it would be cool to see if also ppc64le would hold up 0.6.1. In any case, great to hear you'd be willing to contribute as recipe maintainer! :)

jakirkham commented 5 years ago

Make is required for Linux, at least. Cmake generates makefiles and make is needed to do the build. I think it *was? baked into the anvil image and maybe isn't anymore. I'll look into the full story there. Maybe a # [unix] is more appropriate.

It was. We are moving away from that in conda-forge and towards requiring everything needed for the build (including make).

jayfurmanek commented 5 years ago

For good measure: @conda-forge-admin, please rerender

jayfurmanek commented 5 years ago

No problem with 0.6.1 on ppc64le other than the new MKL test which only makes sense on x86_64, so I added a patch to mark that one as skip for non x86_64 platforms. Still the Windows fail, though.

h-vetinari commented 5 years ago

OK, that's cool. How about you revert this PR to 0.5.0 (sorry for the churn, but you didn't open a separate PR for trying 0.6.1), and then we rebase #12 on that (and can keep trying to fix the windows problems there)? At least then there would be a build for ppc64le already.

jayfurmanek commented 5 years ago

Sounds good. I put up: https://github.com/conda-forge/osqp-feedstock/pull/15