GH-48339: [C++] Enhance functions in util/ubsan.h to support types without a default constructor

[andishgar]

Rationale for this change

arrow::util::SafeLoadAs, arrow::util::SafeLoad, and arrow::util::SafeCopy do not work correctly for types that do not have a default constructor.

What changes are included in this PR?

Changed the behavior of the above-mentioned functions to support types that do not have a default constructor.

Are these changes tested?

Yes.

Are there any user-facing changes?

No.

• GitHub Issue: [C++] Enhance functions in util/ubsan.h to support types that do not have a default constructor #48339

[github-actions]

⚠️ GitHub issue #48339 has been automatically assigned in GitHub to PR creator.

[andishgar]

@pitrou
As I found here, the approach used for bit_cast in this PR is perfectly fine.

Attuned developers use aligned_storage with memcpy, avoiding alignment pitfalls and allowing them to bit-cast non-default-constructible types.

And here is the reference implementation for std::bit_cast.
https://github.com/jfbastien/bit_cast/blob/c288208c1a68a329a5839f007200ab2ee2b65073/bit_cast.h#L57-L59

[andishgar]

@pitrou I’d appreciate your review when you have a moment.

[pitrou]

@github-actions crossbow submit -g cpp

[github-actions]

Revision: dff4b7c

Submitted crossbow builds: ursacomputing/crossbow @ actions-65a6da834d

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-cuda-cpp-ubuntu-24.04-cuda-13.0.2
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

[pitrou]

arrow::util::SafeLoadAs, arrow::util::SafeLoad, and arrow::util::SafeCopy do not work correctly for types that do not have a default constructor.

Is there any type that you have in mind? There are not many types that are trivially constructible but don't have a default constructor, IMHO.

[andishgar]

Is there any type that you have in mind? There are not many types that are trivially constructible but don't have a default constructor, IMHO.

@pitrou
Types with the following pattern are trivially copyable but do not have a default constructor. However, I have not found any such type in Arrow so far.

struct A {
  explicit A(int a) : a(a) {}
  int a;
};

What would you suggest here?

1. Close the PR?
2. Add a separate code path depending on the presence of a default constructor?
3. Or proceed with merging this PR as is?

[pitrou]

I'm ok with merging PR, I was wondering if you had a specific reason for submitting it.

[andishgar]

Mostly for correctness and to avoid relying on an implicit default-constructibility assumption that isn’t required by trivial copyability, even if we don’t currently have such a type in Arrow.

[pitrou]

LGTM, I'll merge

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 3 benchmarking runs that have been run so far on merge-commit aeb6486.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 538 possible false positives for unstable benchmarks that are known to sometimes produce them.