GH-50046: [CI][C++] Improve caching with apache/infrastructure-actions/stash and more general cache keys

[rok]

Rationale for this change

C++ CI jobs currently key ccache/Docker volume caches with hashFiles('cpp/**'). This creates new immutable GitHub Actions cache entries whenever any C++ file changes, even though ccache already handles per-file invalidation internally.

Recent CI logs show that when caches are restored, C++ jobs get high ccache hit rates, but GitHub cache restore misses are common and cause large CI runtime regressions.

What changes are included in this PR?

• Replace actions/cache with apache/infrastructure-actions/stash/restore and apache/infrastructure-actions/stash/save.
• Remove hashFiles('cpp/**') from cache keys.
• Use stable restore prefixes per job/config.

Are these changes tested?

By CI?

Are there any user-facing changes?

CI users should see faster builds..

• GitHub Issue: [CI][C++] Improve ccache performance by using more general cache keys #50046

[github-actions]

⚠️ GitHub issue #50046 has been automatically assigned in GitHub to PR creator.

[Copilot]

Pull request overview

This PR refactors C++ CI cache handling to improve ccache reuse. Instead of using actions/cache with cache keys that include hashFiles('cpp/**'), it splits cache restore/save into separate steps, uses github.run_id as the unique save key, and only saves new caches when running on main. This avoids creating new cache entries per PR (which can cause eviction) while still letting PR builds benefit from the stable restore prefix.

Changes:

• Replace actions/cache@v5 with paired actions/cache/restore@v5 and actions/cache/save@v5 steps across C++ workflows.
• Drop hashFiles('cpp/**') from cache keys; use ${{ github.run_id }} as the save key and a stable prefix as the restore-key.
• Gate Save steps with if: github.ref == 'refs/heads/main' so only main pushes write new cache entries.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
.github/workflows/cpp.yml	Splits Docker volume and ccache caching into restore/save pairs for docker, macOS, and MinGW jobs; saves only from main.
.github/workflows/cpp_windows.yml	Splits Windows ccache caching into restore/save; saves only from main.
.github/workflows/cpp_extra.yml	Same restore/save split applied to docker-extra, JNI (linux + macOS), ODBC linux/macOS/MSVC jobs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

[kou]

https://github.com/apache/infrastructure-actions/tree/main/stash
apache/infrastructure-actions has a similar action that is originally developed by @assignUser.
Should we try this?

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

[rok]

@kou I think just tweaking ccache could have worked, but infrastructure-actions/stash gives us 5 day eviction and no size limit so a safer long term approach.

I updated the PR to use infrastructure-actions/stash. I am not sure if there's another way to test it but by mergeing?

[kou]

Could you update the PR title and description?

Let's try this. If this doesn't work, we can fix this or revert this later.

[pitrou]

If we switch to apache/infrastructure-actions/stash, do we still need to restrict cache saves to git main?

[rok]

Thanks for the review @kou & @pitrou. I updated the PR title/description and allowed for cache to be stored on non-main branches as well.

[rok]

@pitrou any further comments or shall we merge and see what happens?

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.