BUILD-10739 Cache cleanup: key-only delete + job summary

[mikolaj-matuszny-ext-sonarsource]

Summary

Two additions to the cache cleanup script:

• Delete by key across all branches: Providing only key (no branch) now enters delete mode instead of list-only. List mode only activates when both branch and key are empty.
• GitHub Actions job summary: Writes a markdown summary to $GITHUB_STEP_SUMMARY with a collapsible <details> table of matched objects. Works in list, dry-run, and delete modes. No-op when running locally.

Updated modes of operation

Scenario	Branch	Key	Dry-run
List all cache entries	(empty)	(empty)	n/a
Preview what would be deleted	feature/my-branch	(optional)	true
Delete cache for a branch	feature/my-branch	(optional)	false
Delete key across all branches	(empty)	sccache-Linux-	false

Test plan

• Test list mode locally (no branch, no key) - still works
• Test key-only delete with dry-run
• Verify GITHUB_STEP_SUMMARY renders correctly in Actions UI (collapsible table)
• Verify script works locally without GITHUB_STEP_SUMMARY set

🤖 Generated with Claude Code

[sonar-review-alpha]

Summary

This PR adds two features to the cache cleanup script:

1. Key-only deletion — the script now supports deleting a cache key across all branches. The list/delete mode logic changed: list mode only activates when both branch and key are empty; providing just a key now triggers delete mode. This enables cleanup scenarios like "delete all sccache-Linux-* across the entire repo."

2. GitHub Actions job summary — a new write_summary() function writes markdown output to $GITHUB_STEP_SUMMARY with object count, total size, and a collapsible table of matched keys. It writes in all three modes (list, dry-run, delete) and safely no-ops when running locally.

Supporting changes: refactored object row extraction into a shared extract_rows() helper to avoid duplicating the jq transform between table output and the summary.

What reviewers should know

Start with: the mode detection logic change at line 175 — this is the crux. Compare the old [[ -z "$INPUT_BRANCH" ]] check against the new [[ -z "$INPUT_BRANCH" && -z "$KEY_PATTERN" ]] to understand the three-way branch/key/both logic.

Watch for: the extract_rows() refactor. It's now the single source of truth for turning S3 object JSON into TSV rows — verify it's called correctly by both print_table() and write_summary(), and that TSV field order (size, date, key) matches expectations in both consumers.

GitHub Actions integration: write_summary() only writes if $GITHUB_STEP_SUMMARY exists (safe locally). Review the markdown table format — it needs to render cleanly in the Actions UI with the collapsible <details> wrapper.

Test the boundary cases:

• Empty branch + empty key → list mode (unchanged behavior)
• Empty branch + key-only → delete mode across all branches (new)
• Branch + empty key → delete for that branch (unchanged)
• All three should produce job summaries in Actions

---

• Generate Walkthrough
• Generate Diagram

🗣️ Give feedback

[sonar-review-alpha]

Conclusion: The feature logic and mode-switching are correct, but write_summary has a formatting bug that will surface immediately in any real run, and it duplicates extraction logic from print_table in a way that will cause silent drift over time.

🗣️ Give feedback

[sonarqube-cloud-us]

SonarQube reviewer guide

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[sonarqubecloud]

SonarQube reviewer guide

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[sonar-review-alpha]

LGTM! ✅

🗣️ Give feedback