Patch Arrow to set CMAKE_POLICY_VERSION_MINIMUM for RapidJSON

[KyleFromNVIDIA]

Description

Provide a patch for apache/arrow#48801

Fixes #22540

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[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: 2ae42f20-5f38-4ee8-b1e4-bf5a423da526

📥 Commits

Reviewing files that changed from the base of the PR and between 2021624 and 351bcdf.

📒 Files selected for processing (1)

• cpp/cmake/thirdparty/patches/override.json

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

• cpp/cmake/thirdparty/patches/override.json

---

📝 Walkthrough

Summary by CodeRabbit

• Chores
  • Apply an override to pin the Arrow dependency and include a compatibility patch to limit RapidJSON’s CMake policy impact during discovery.
  • Update build initialization to automatically load the dependency override file, improving consistency and reproducibility of builds.

Walkthrough

This PR scopes a temporary CMake policy change to CMAKE_POLICY_VERSION_MINIMUM=3.5 around RapidJSON discovery, registers that patch in an Arrow override JSON, and passes the override to rapids_cpm_init during native build initialization.

Changes

CMake RapidJSON Policy Configuration

Layer / File(s)	Summary
RapidJSON CMake policy patch cpp/cmake/thirdparty/patches/arrow_rapidjson_cmake_policy_version_minimum.diff	CMake patch saves and restores CMAKE_POLICY_VERSION_MINIMUM around the find_package(RapidJSON ...) call, limiting policy-minimum scope to that invocation.
Arrow dependency override configuration cpp/cmake/thirdparty/patches/override.json	JSON configuration defines Arrow 21.0.0 from Apache Arrow repository with full clone and applies the RapidJSON CMake policy patch.
Build system integration java/src/main/native/CMakeLists.txt	Rapids CPM initialization switches from bare rapids_cpm_init() to parameterized call with OVERRIDE pointing to the patch override configuration file.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related issues

• [BUG] Java JNI build fails with CMake 4.0.7 due to Arrow/RapidJSON cmake_minimum_required compatibility #22540: The temporary CMAKE_POLICY_VERSION_MINIMUM=3.5 setting in the RapidJSON patch directly addresses the CMake 4 compatibility failure described in this issue.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title directly describes the main change: patching Arrow to set CMAKE_POLICY_VERSION_MINIMUM for RapidJSON, which aligns with the changeset's primary objective.
Description check	✅ Passed	The description references issue #22540 which documents the Arrow/RapidJSON CMake compatibility problem that this PR addresses through the provided patch.
Linked Issues check	✅ Passed	The PR successfully implements the required workaround: it creates a patch file for Arrow that sets CMAKE_POLICY_VERSION_MINIMUM to 3.5 during RapidJSON's find_package call, directly addressing the CMake 4.0.7 compatibility issue [#22540].
Out of Scope Changes check	✅ Passed	All changes are narrowly scoped to implementing the required Arrow patch and applying it via the override.json configuration and CMakeLists.txt update, with no extraneous modifications.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

cpp/cmake/thirdparty/patches/override.json (1)

8-8: ⚡ Quick win

Consider using shallow clone if patch target is available at the specified tag.

Setting git_shallow: false forces a full clone of the Arrow repository, which can significantly increase initial build times (Arrow repo is large). Since the patch targets cpp/cmake_modules/FindRapidJSONAlt.cmake — a core CMake infrastructure file that exists in the apache-arrow-21.0.0 tag — a shallow clone would be sufficient for applying this patch. Shallow clones are preferred unless the patch requires commit history or targets files not present in the specified tag.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@cpp/cmake/thirdparty/patches/override.json` at line 8, The override JSON
currently sets "git_shallow": false causing full clones of the Arrow repo;
change it to "git_shallow": true in cpp/cmake/thirdparty/patches/override.json
(the JSON key "git_shallow") so the fetch uses a shallow clone for the
apache-arrow-21.0.0 tag (since the patched file
cpp/cmake_modules/FindRapidJSONAlt.cmake exists in that tag) unless the patch
requires history—ensure no other build step needs full history before flipping
the flag.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In
`@cpp/cmake/thirdparty/patches/arrow_rapidjson_cmake_policy_version_minimum.diff`:
- Around line 9-12: The current patch unconditionally saves and restores
CMAKE_POLICY_VERSION_MINIMUM which causes an empty string to be written if the
variable was originally unset; update the logic around the save/restore of
CMAKE_POLICY_VERSION_MINIMUM (the temporary variable
_CMAKE_POLICY_VERSION_MINIMUM_OLD and the restore after find_package(RapidJSON
...)) to first test whether CMAKE_POLICY_VERSION_MINIMUM was defined (using
CMake's DEFINED check), only set _CMAKE_POLICY_VERSION_MINIMUM_OLD when it
exists, set CMAKE_POLICY_VERSION_MINIMUM to 3.5 for the RapidJSON find, and then
restore the previous state by either resetting CMAKE_POLICY_VERSION_MINIMUM to
the saved value or unsetting it when it was originally undefined.

---

Nitpick comments:
In `@cpp/cmake/thirdparty/patches/override.json`:
- Line 8: The override JSON currently sets "git_shallow": false causing full
clones of the Arrow repo; change it to "git_shallow": true in
cpp/cmake/thirdparty/patches/override.json (the JSON key "git_shallow") so the
fetch uses a shallow clone for the apache-arrow-21.0.0 tag (since the patched
file cpp/cmake_modules/FindRapidJSONAlt.cmake exists in that tag) unless the
patch requires history—ensure no other build step needs full history before
flipping the flag.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 5ae8235f-e502-41a2-b859-ca0a53609a07

📥 Commits

Reviewing files that changed from the base of the PR and between 2dba3ec and 85fc85d.

📒 Files selected for processing (3)

• cpp/cmake/thirdparty/patches/arrow_rapidjson_cmake_policy_version_minimum.diff
• cpp/cmake/thirdparty/patches/override.json
• java/src/main/native/CMakeLists.txt

[pmattione-nvidia]

This is just for the release branch. Does it need to go into the main branch as well?

[KyleFromNVIDIA]

It will be automatically forward-merged into the main branch, no problem.

[KyleFromNVIDIA]

/merge