GH-49921: [C++] Bump xsimd to 14.2.0

[AntoinePrv]

Rationale for this change

• Fix for the current SVE ODR violation in the latest version
  Fix SVE dynamic dispatch ODR violation xtensor-stack/xsimd#1311
• Pre-requisit to [C++] Use existing library for CPU detection #45331

What changes are included in this PR?

Bump version and cleanup backports.

Are these changes tested?

Yes, in CI.

Are there any user-facing changes?

No

• GitHub Issue: [C++] Bump xsimd to 14.2.0 #49921

[github-actions]

⚠️ GitHub issue #49921 has been automatically assigned in GitHub to PR creator.

[pitrou]

@AntoinePrv Can you take a look at the CI failures? They seem related.

[AntoinePrv]

Yes, I am still waiting for the Msys2 build of xsimd to happen...

[pitrou]

There are failures on non-MSys2 builds, though:

• https://github.com/apache/arrow/actions/runs/25421952127/job/74565986944?pr=49922
• https://github.com/apache/arrow/actions/runs/25421952024/job/74565986771?pr=49922

[AntoinePrv]

Perhaps it's this check that does not work on MSVC?

https://github.com/xtensor-stack/xsimd/blob/80c23624ce008d937da7e845e528e82ce0cbf4e0/include/xsimd/types/xsimd_batch_constant.hpp#L500

Either we could add /Zc:__cplusplus or re-introduce the backport.

[pitrou]

I think we should both 1) fix xsimd upstream anyway and 2) pass /Zc:__cplusplus when building Arrow C++.

It seems xsimd should do something like this:

#ifdef _MSVC_LANG
#define XSIMD_CPP_LEVEL _MSVC_LANG
#else
#define XSIMD_CPP_LEVEL __cplusplus
#endif

and then use the XSIMD_CPP_LEVEL macro everywhere instead of __cplusplus.

(but xsimd could also use find-grained feature testing).

[AntoinePrv]

Already sending a patch, but not sure I want to go through another round of xsimd releases

xtensor-stack/xsimd#1342

[pitrou]

Agreed, hence "pass /Zc:__cplusplus when building Arrow C++" as well 😉

[pitrou]

Bloody protobuf fails building with /Zc:__cplusplus.
https://github.com/apache/arrow/actions/runs/25429128655/job/74590408553?pr=49922#step:9:2218

[pitrou]

Tiny syntax error that needs fixing:

  #17 27.36 Failed to load manifest from directory /arrow/ci/vcpkg
  #17 27.36 /arrow/ci/vcpkg/vcpkg.json:24:26: error: Trailing comma in an object
  #17 27.36   on expression:       "version>=": "14.2",

[AntoinePrv]

@pitrou I'm not very proactively working on this until the msys2 xsimd build is live

[pitrou]

@pitrou I'm not very proactively working on this until the msys2 xsimd build is live

@kou Do you know how we can make that happen?

[pitrou]

@github-actions crossbow submit -g cpp

[github-actions]

Revision: 89a1f7e

Submitted crossbow builds: ursacomputing/crossbow @ actions-9c434593fe

Task	Status
example-cpp-minimal-build-static
example-cpp-minimal-build-static-system-dependency
example-cpp-tutorial
test-build-cpp-fuzz
test-conda-cpp
test-conda-cpp-valgrind
test-debian-13-cpp-amd64
test-debian-13-cpp-i386
test-debian-experimental-cpp-gcc-15
test-fedora-42-cpp
test-ubuntu-22.04-cpp
test-ubuntu-22.04-cpp-bundled
test-ubuntu-22.04-cpp-emscripten
test-ubuntu-22.04-cpp-no-threading
test-ubuntu-24.04-cpp
test-ubuntu-24.04-cpp-bundled-offline
test-ubuntu-24.04-cpp-gcc-13-bundled
test-ubuntu-24.04-cpp-gcc-14
test-ubuntu-24.04-cpp-minimal-with-formats
test-ubuntu-24.04-cpp-thread-sanitizer

[raulcd]

Do you know how we can make that happen?

It's just a matter of waiting for their queue to update it:
https://packages.msys2.org/queue

Based on experience when uploading Arrow it takes between 6 and 12 hours. Their CI seems to run several times a day and xsimd is already listed on the queue so it should be available "soon".

[raulcd]

Sorry, just thought about this, I am happy to merge as is.
Are we planning to bump this requirement later on when we improve the cpu detection feature?

arrow/cpp/cmake_modules/ThirdpartyToolchain.cmake

Lines 2855 to 2861 in 3d4ab3f

	resolve_dependency(xsimd
	FORCE_ANY_NEWER_VERSION
	TRUE
	IS_RUNTIME_DEPENDENCY
	FALSE
	REQUIRED_VERSION
	"14.0.0")

[raulcd]

@github-actions crossbow submit wheelcp314

[github-actions]

Revision: 8278222

Submitted crossbow builds: ursacomputing/crossbow @ actions-0d5b743af7

Task	Status
wheel-macos-monterey-cp314-cp314-amd64
wheel-macos-monterey-cp314-cp314-arm64
wheel-macos-monterey-cp314-cp314t-amd64
wheel-macos-monterey-cp314-cp314t-arm64
wheel-manylinux-2-28-cp314-cp314-amd64
wheel-manylinux-2-28-cp314-cp314-arm64
wheel-manylinux-2-28-cp314-cp314t-amd64
wheel-manylinux-2-28-cp314-cp314t-arm64
wheel-musllinux-1-2-cp314-cp314-amd64
wheel-musllinux-1-2-cp314-cp314-arm64
wheel-musllinux-1-2-cp314-cp314t-amd64
wheel-musllinux-1-2-cp314-cp314t-arm64
wheel-windows-cp314-cp314-amd64
wheel-windows-cp314-cp314t-amd64

[raulcd]

The C GLib & Ruby failure on Windows is related. I think we also have to update the vcpkg.json baseline version here:

arrow/cpp/vcpkg.json

Line 60 in 3d4ab3f

"builtin-baseline": "40c89449f0ccce12d21f8a906639f6c2c649b9e7"

and here:

arrow/c_glib/vcpkg.json

Line 10 in 3d4ab3f

"builtin-baseline": "40c89449f0ccce12d21f8a906639f6c2c649b9e7"

We should update to the same one we have on .env (9b965a116838c6cdcd36bca60d1b81b030c8ab8d), sorry I did not realised about that. Can you give that a quick try. If it fails due to any other build failure, I'll try to bump it on isolation and fix any possible CI issue.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

[raulcd]

ok, the latest fixed the CI failure. Current CI failures are unrelated, there's still some CI jobs finishing but if no surprises, I'll merge once they finish

[AntoinePrv]

Thanks @raulcd ! Looks like this is ready now

[raulcd]

Merged! Thanks @AntoinePrv for following it up!

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit a10f254.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 107 possible false positives for unstable benchmarks that are known to sometimes produce them.