GH-47483: [C++] Bump vendored xxhash to 0.8.3

[raulcd]

Rationale for this change

We are using an older version, this might fix the current failure for test-conda-cpp-valgrind failures. The upgrade might also come with some improvements / fixes.

What changes are included in this PR?

Update vendored xxhash to newer version:

• https://github.com/Cyan4973/xxHash/releases/tag/v0.8.3

Are these changes tested?

Yes, via CI.

Are there any user-facing changes?

No

• GitHub Issue: [C++] Bump vendored version of xxhash to latest #47483

[github-actions]

⚠️ GitHub issue #47475has been automatically assigned in GitHub to PR creator.

[raulcd]

Current failures are related to LLVM, I will rebase once the PR fixing those jobs is merged.

[raulcd]

@github-actions crossbow submit -g cpp

[github-actions]

Revision: 29e3ab6

Submitted crossbow builds: ursacomputing/crossbow @ actions-9f25a8e327

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-cuda-cpp-ubuntu-22.04-cuda-11.7.1
test-debian-12-cpp-amd64
test-debian-12-cpp-i386
test-fedora-42-cpp
test-ubuntu-22.04-cpp
test-ubuntu-22.04-cpp-20
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]

It seems that the xxhash bump doesn't fix the test-conda-cpp-valgrind issues:
https://github.com/ursacomputing/crossbow/actions/runs/17428470869/job/49481058706

Still, seems worth updating it, @pitrou what do you think?
Should I create an issue for the version bump and we handle the job failure on the original issue?

[pitrou]

Yes, it's worth bumping xxhash anyway.

[github-actions]

⚠️ GitHub issue #47483has been automatically assigned in GitHub to PR creator.

[pitrou]

It appears this was reported and discussed upstream: Cyan4973/xxHash#943 (comment)

We should probably apply the stated workaround:

the only thing I can suggest at this point is to recommend using XXH_NO_INLINE_HINTS when compiling with -Og

[raulcd]

@github-actions crossbow submit test-conda-cpp-valgrind

[github-actions]

Revision: eb76ca8

Submitted crossbow builds: ursacomputing/crossbow @ actions-d2be094bc2

Task	Status
test-conda-cpp-valgrind

[pitrou]

@kou Does it look like the right way to do this?

[kou]

No.

In general, we should not use add_definitions() because its scope is global. We should create a target and use it instead:

add_library(arrow::xxhashINTERFACEIMPORTED)
target_compile_definitions(arrow::xxhashINTERFACE"$<$<CONFIG:Debug>:XXH_NO_INLINE_HINTS>")

(I think that we don't need -Og check. We can always disable inline hints in debug build.)

[kou]

FYI: #47495 is an approach that doesn't use add_definitions().

[raulcd]

Thanks @kou , should we create targets for the different vendored libraries and only link when necessary as a future task too? I am thinking about datetime, double-conversion, etc

[kou]

Yes when they need additional build flags. If they don't need additional build flags, we don't need to create targets for them. Because we have src/ in include paths:

arrow/cpp/CMakeLists.txt

Lines 559 to 560 in 743d0af

	include_directories(${CMAKE_CURRENT_BINARY_DIR}/src)
	include_directories(src)

BTW, we should not use include_directories() because its scope is global (directory) too...

[raulcd]

The CI failure is now unrelated to building but to a leak on azure-test.

The following tests FAILED:
77 - arrow-azurefs-test (Failed)

[pitrou]

Cool, I suspect we can merge but let's just first wait for AppVeyor to finish?

[pitrou]

I restarted the Valgrind job. If the Azure-related failure occurs again, we'll have to file a separate issue for it.

[raulcd]

I restarted the Valgrind job. If the Azure-related failure occurs again, we'll have to file a separate issue for it.

It has failed again, I've opened a new issue to track the Azure leak:

• [C++] Valgrind fails consistently on Azure tests #47485

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 2284234.

There weren't enough matching historic benchmark results to make a call on whether there were regressions.

The full Conbench report has more details.