[CI] Mac jar build uses a different LLVM version than Linux
- Dominant language
- Java
- Stars
- 94
- Forks
- 152
- Avg merge
- 3d 16h
- Merged PRs (30d)
- 11
Description
### Describe the bug, including details regarding any error messages, version, and platform.
The Mac build uses an older LLVM version (15) than Linux (18). This seems to be related to the CI code and the fact that the mac github runners have llvm installed via brew
https://github.com/actions/runner-images/blob/releases/macos-14-arm64/20250915/images/macos/macos-14-arm64-Readme.md
Even newer runners only have LLVM 17.
The discrepancy could lead to hard to find bugs or performance issues.
Is it possible to use vcpkg for the mac builds? That way we could be assured that they are always consistent across platforms.
A few related changes I found:
Unpin the llvm version in arrow/cpp/Brewfile. I think this might mean that theoretically the latest llvm could be installed but since the runner image already has an older one installed, that is used instead.
https://github.com/apache/arrow/commit/c1f995226c46c2bba9c5fbf988cf9b247f87adad
A related change talking about llvm 14. https://github.com/apache/arrow/pull/14310
Contributor guide
Research direction
Locate the macOS and Linux jar-build CI definitions and compare how LLVM is selected, then inspect arrow/cpp/Brewfile and the linked runner-image details. Done should mean the builds use a consistent LLVM version or have an explicitly documented, justified version difference across platforms.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- github-actions, java
- Domain
- build-system, ci-cd
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 38/100