[libcu++] Fix cuda::std::aligned_alloc argument order

[davebayer]

This is bad, we had the argument order swapped..

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: c541988e-b3ea-422a-bfe1-b683b0f62d96

📥 Commits

Reviewing files that changed from the base of the PR and between e0a0e23 and 8cf758a.

📒 Files selected for processing (3)

• c/parallel.v2/src/hostjit/include/hostjit/cuda_minimal/stubs/cuda/std/__cstdlib/aligned_alloc.h
• libcudacxx/include/cuda/std/__cstdlib/aligned_alloc.h
• libcudacxx/test/libcudacxx/std/language.support/cstdlib/aligned_alloc.pass.cpp

🚧 Files skipped from review as they are similar to previous changes (1)

• libcudacxx/include/cuda/std/__cstdlib/aligned_alloc.h

---

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes
  • Corrected aligned_alloc parameter ordering so the alignment argument is taken before the size argument, consistently across host and device code paths.
• Tests
  • Updated aligned allocation tests to use the new aligned_alloc argument order and corresponding noexcept expectations.

Walkthrough

aligned_alloc and __aligned_alloc_host now take (align, nbytes) instead of (nbytes, align). The host forwarding path and stub declarations were updated, and the aligned_alloc test now calls the reordered API.

Changes

Aligned Allocation Parameter Reordering

Layer / File(s)	Summary
Reorder API and host forwarding libcudacxx/include/cuda/std/__cstdlib/aligned_alloc.h, c/parallel.v2/src/hostjit/include/hostjit/cuda_minimal/stubs/cuda/std/__cstdlib/aligned_alloc.h	The public wrapper, host helper, and stub declarations use (align, nbytes) ordering; the host branch forwards the reordered arguments, and the device syscall declaration is given named parameters.
Update aligned_alloc test libcudacxx/test/libcudacxx/std/language.support/cstdlib/aligned_alloc.pass.cpp	The test now invokes cuda::std::aligned_alloc with (align, size) and updates the matching noexcept check.

important: this changes a public parameter order on aligned_alloc; any positional callers or bindings still using (nbytes, align) will pass swapped values.

Suggested reviewers: miscco, bernhardmgruber

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 2

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 0c8a2c20-16df-4c9b-9b0d-7af370e74f35

📥 Commits

Reviewing files that changed from the base of the PR and between 886c291 and e0a0e23.

📒 Files selected for processing (1)

• libcudacxx/include/cuda/std/__cstdlib/aligned_alloc.h

[github-actions]

😬 CI Workflow Results

🟥 Finished in 1h 57m: Pass: 97%/120 | Total: 3d 07h | Max: 1h 56m | Hits: 54%/857433

See results here.

[bernhardmgruber]

So, we introduced this in #3339, right? So we are shipping this bug since CCCL 3.0?