build: apply patch on Arrow so it can be used with older versions of CMake

[raulcd]

For older versions of CMake the variable arrow_SOURCE_DIR is overriden by FetchContent's one because project(arrow) sets it as a cache variable while FetchContent sets a normal variable with precedence. Arrow can't find thirdparty/versions.txt because it is pointing to the tarball root instead of the cpp subfolder. See policy https://cmake.org/cmake/help/latest/policy/CMP0180.html

We should fix upstream Arrow so it does not rely on arrow_SOURCE_DIR to allow older CMake builds.

Fixes #1028

[raulcd]

@wgtmac I have tested this locally against a Dockerfile with 24.04 and CMake 3.28.3. As we seem to require cmake_minimum_required(VERSION 3.25) we might want to add a test for a lower CMake version?

I've opened the following issue to follow up on Arrow:

• [C++] Building Arrow via FetchContent can fail with versions older than CMake 3.30 arrow#53995

[Copilot]

🟡 Changes recommended

Tool validation can reject installed-Arrow builds, and the Git fallback can silently leave vendored Arrow unpatched.

2 open findings

• Ensure git apply patches Arrow outside the enclosing repository · New
• Require patch tools only when vendoring Arrow · New

What changed in this PR

Fixes the vendored Arrow configuration failure on older CMake versions described in #1028.

Changes:

• Adds an Arrow patch using ARROW_SOURCE_DIR for third-party paths.
• Integrates patch application into FetchContent, with a Git fallback.

File	Description
cmake_modules/​IcebergThirdpartyToolchain.cmake	Adds patch-tool selection and FetchContent patching.
cmake_modules/​arrow-thirdparty-dir.patch	Corrects Arrow’s third-party source directory reference.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[raulcd]

This is basically the same approach we take at Arrow, I don't think it's true, I've applied with both git apply and patch.

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

2 open findings

• PATCH_COMMAND ${ARROW_PATCH_COMMAND} is always emitted, but ARROW_PATCH_COMMAND can be empty… · New
• Ensure git apply patches Arrow outside the enclosing repository

1 resolved since last review

• Require patch tools only when vendoring Arrow

🧠 Review effort: Lite