xsimd: 13.2.0 -> 14.2.0

[doronbehar]

Things done

• Built on platform:
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• Tested, as applicable:
  • NixOS tests in nixos/tests.
  • Package tests at passthru.tests.
  • Tests in lib/tests or pkgs/test for functions and "core" functionality.
• Ran nixpkgs-review on this PR. See nixpkgs-review usage.
• Tested basic functionality of all binary files, usually in ./result/bin/.
• Nixpkgs Release Notes
  • Package update: when the change is major or breaking.
• NixOS Release Notes
  • Module addition: when adding a new NixOS module.
  • Module update: when the change is significant.
• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.

---

Add a 👍 reaction to pull requests you find important.

[K900]

Have you managed to build it on aarch64-linux? I got a bunch of test failures consistently.

[doronbehar]

Me too.. (When the PR is rebased upon branch master). Here's an upstream bug report:

• Many test failures on aarch64-linux xtensor-stack/xsimd#1232

Since I doubt they will reply soon, and also because these 42 failures are pretty severe for a platform we wish to officially support, I'm marking this as a draft.

[doronbehar]

After someone has bisected the upstream commit that broke the tests on aarch64-linux, I reverted that upstream patch and it builds fine now (that's what I just pushed). I'm keeping this PR marked as a draft because I believe upstream will want to make an effort to fix this a bit more properly. The branch of this PR is still available in case someone strongly needs the latest and greatest xsimd 14.0.0 on aarch64-linux.

[sifmelcara]

Is this still waiting for upstream fixes given that 14.1.0 has been out for a month?

[doronbehar]

Is this still waiting for upstream fixes given that 14.1.0 has been out for a month?

The PR suggests to update xsimd to 14.1.0 at the moment, but it doesn't fix the upstream issue, see xtensor-stack/xsimd#1232 . I'm still hesitent as for the revert.

If you are very eager to make this version bump happen, maybe try to build on aarch64-linux a few of the direct dependencies of xsimd. Try to do that on branch master to reduce the amount of required builds to a minimum.

[sifmelcara]

The PR suggests to update xsimd to 14.1.0 at the moment, but it doesn't fix the upstream issue, see xtensor-stack/xsimd#1232 . I'm still hesitent as for the revert.

If you are very eager to make this version bump happen, maybe try to build on aarch64-linux a few of the direct dependencies of xsimd. Try to do that on branch master to reduce the amount of required builds to a minimum.

Thanks I leaved a comment in the upstream ticket. My concern is that this doesn't seem to get enough attention from upstream maintainer so it might not get fixed anytime soon. Given that many other distro have updated to 14, so unsure if it actually makes sense to hold off the update in nixpkgs. If the bug have little practical impact on downstream packages, it should maybe be considered as an upstream bug and shouldn't block the update here.

Unfortunately I don't have aarch64 machine to test

[doronbehar]

@sifmelcara this should be ready now - I tested the build of xsimd on both aarch64-{linux,darwin}.

[doronbehar]

I'll merge this on Tuesday, to allow people to review this one last time.

[sifmelcara]

Would we also need to patch xtensor-stack/xsimd@233b2de to make sure aarch64 test passes in nixpkgs?

[doronbehar]

Would we also need to patch xtensor-stack/xsimd@233b2de to make sure aarch64 test passes in nixpkgs?

The current fetchpatch call refers to the merge commit of upstream's xtensor-stack/xsimd#1241 PR which includes the upstream commit xtensor-stack/xsimd@233b2de . Note that I tested the current state of the PR merge upon branch master, also on aarch64-linux and aarch64-darwin.

[sifmelcara]

The current fetchpatch call refers to the merge commit of upstream's xtensor-stack/xsimd#1241 PR which includes the upstream commit xtensor-stack/xsimd@233b2de . Note that I tested the current state of the PR merge upon branch master, also on aarch64-linux and aarch64-darwin.

Maybe I'm blind but I don't that's true? The hash isn't referring to a merge commit, and when I open the URL (https://github.com/xtensor-stack/xsimd/commit/0b97964a6561eba91ca07fefbeb19196cab7cb44.patch) to check the patch content, it doesn't include xtensor-stack/xsimd@233b2de

[sifmelcara]

The commits in this PR also aren't squashed and cleaned up, maybe you simply forgot to push your local changes?

[doronbehar]

You are right, it wasn't a merge commit I fetchpatched.. Anyway, evidently the aarch64-linux and aarch64-darwin issues are both fixed.

As for the commit separation, I kind of like it how things are separated now, but we can wait for a new upstream release as it should reach us soon, according to them.

[doronbehar]

On branch master, on all 4 platforms, I tested the build of: xsimd 14.2.0 and tested that scipy (directly dependent package) builds with it. All builds were successful. Hence I'm merging.

I decided to do squash some of the changes to a single version bump. The details of the state of the PR in between will probably won't be of any benefit to anyone.