CI adjustments for conda-libmamba-solver as default

[jaimergp]

Description

conda 23.10.0 requires conda-libmamba-solver and has solver=libmamba as default, so conda-build's CI is going to be using it now.

Checklist - did you ...

• Add a file to the news directory (using the template) for the next release's release notes?
• Add / update necessary tests?
• Add / update outdated documentation?

[mbargull]

LGTM

[mbargull]

@jaimergp, this is still marked as a draft.
Any reason not to merge?

[jaimergp]

Just the pre-commit and missing news. I'll commit them now. Thanks for the reminder!

[jaimergp]

Errors are related to conda-libmamba-solver itself, probably to be fixed by conda/conda-libmamba-solver#365.

[jaimergp]

Let's see how conda-libmamba-solver 23.11.1 does now :)

[kenodegard]

@jaimergp no idea why tests/cli/test_main_build.py::test_build_output_folder and tests/cli/test_main_build.py::test_build_source are failing (see https://github.com/conda/conda-build/actions/runs/6897842353/job/18770690271?pr=5059#step:8:3164), attempting to fix it by switching to tmp_path, see 031dac4

[jaimergp]

Me neither @kenodegard 🤷 Thanks for the help though! I'm going to be the solver-related issue tests/test_api_build.py::test_build_msvc_compiler[14.0] - KeyError: Dist(channel='local', dist_name='conda-build-test-build-msvc-compiler-1.0-0', name='conda-build-test-build-msvc-compiler', fmt='.tar.bz2', version='1.0', build_string='0', build_number=0, base_url=None, platform=None) in conda/conda-libmamba-solver#386. The other things might be artifacts of the solver now being faster, which means things need to be recreated more quickly than before? Idk. If feels like a race condition.

[mbargull]

@jaimergp no idea why tests/cli/test_main_build.py::test_build_output_folder and tests/cli/test_main_build.py::test_build_source are failing (see https://github.com/conda/conda-build/actions/runs/6897842353/job/18770690271?pr=5059#step:8:3164), attempting to fix it by switching to tmp_path, see 031dac4

This is most likely fixed by gh-5068 .
What we have here is that the croot is being modified and/or removed by concurrently running tests which is why conda-index fails.

[mbargull]

It would be good if we could merge this PR and gh-5068 to remove some of the failing test clutter.
Further fixes should be done/investigated in subsequent PRs.
@jaimergp, mark as "Ready for review"?

[mbargull]

@jaimergp no idea why tests/cli/test_main_build.py::test_build_output_folder and tests/cli/test_main_build.py::test_build_source are failing (see https://github.com/conda/conda-build/actions/runs/6897842353/job/18770690271?pr=5059#step:8:3164), attempting to fix it by switching to tmp_path, see 031dac4

This is most likely fixed by gh-5068 . What we have here is that the croot is being modified and/or removed by concurrently running tests which is why conda-index fails.

Ah, taking another look at those 2 tests, I see that they use --croot explicitly; so those might not necessarily be affected by gh-5068. (We should merge that in nonetheless.)

[jaimergp]

Agreed @mbargull. Thanks!

[jezdez]

@mbargull Just approved, but I wanted to clarify that it's a-ok to merge given the new build tools team effort. 🚀

[mbargull]

@mbargull Just approved, but I wanted to clarify that it's a-ok to merge given the new build tools team effort. 🚀

❤️ Thanks.
(I've been out for so long, I'm only slowly getting the lay of the land of the current state of who/what/when in Conda land ;).)

[dholth]

That's too bad.

[jezdez]

libmamba doesn't support unicode package names IIUC, no need to test for it.

[kenodegard]

Merged in #5068, any remaining changes necessary to get everything back to 🟢 should be added to this PR