Deflake release-readiness recency test: remove wall-clock-dependent live assertions, add deterministic fixture coverage

[PureWeen]

Note

Are you waiting for the changes in this PR to be merged?
It would be very helpful if you could test the resulting artifacts from this PR and let us know in a comment if this change resolves your issue. Thank you!

Summary

Follow-up to #35971. De-flakes a pre-existing time-bomb in the release-readiness skill's own test suite (.github/skills/release-readiness/tests/Test-ReleaseReadiness.ps1). Skills-only change — no framework code touched.

The time-bomb

The suite had four live, wall-clock-dependent assertions that hardcoded hasRecentActivity = $true against real release branches (SR8, SR9, the active-SR foreach loop, and preview6).

The detector (Find-ReleaseReadinessTrackers.ps1) computes hasRecentActivity from Get-RecentCommitCount, which runs git log <branch> --since=7.days and returns (count > 0). So those assertions are only true while the real branch has had a commit in the last 7 days. The moment a servicing branch goes quiet for 7 days — a normal end-of-cycle state — the assertion flips red.

On 2026-06-18 this actually happened: SR8's last commit (the SR7→SR8 merge #35810) landed 2026-06-11, so --since=7.days returned 0 and the two SR8 assertions went red. They self-heal on the next commit, but nondeterministic red in a suite whose entire selling point is determinism is a credibility bug.

What changed

• Removed the wall-clock-dependent live assertions. The end-to-end detector run now asserts only that hasRecentActivity is a real [bool] the detector emitted — never a date-dependent value.
• Added genuinely deterministic coverage of the recency-window math via a synthetic fixture: a throwaway temp git repo with commits at controlled dates (GIT_AUTHOR_DATE/GIT_COMMITTER_DATE at now−6d / now−8d / now−30d), then calls the real Get-RecentCommitCount (dot-sourced) against it and asserts exact counts for the 7/10/60/1-day windows plus the origin/-prefixed ref form. Zero network, zero dependence on "today"; the temp repo is cleaned up in a finally.
• Corrected the misleading comments that equated "active SR" with hasRecentActivity = true. An active SR can legitimately idle for >7 days; hasRecentActivity is a 7-day-window signal, not a synonym for "active".

Verification

Full suite is green: pwsh -NoProfile -File .github/skills/release-readiness/tests/Test-ReleaseReadiness.ps1 → Passed: 552, Failed: 0 (was 545/2 before, with the two SR8 reds). The new fixture assertions also pass under -SkipE2E (offline), confirming they're network-independent.

Note: the maui-pr framework pipeline intentionally skips for skills-only PRs — that's expected, not a failure.

[github-actions]

🚀 Dogfood this PR with:

⚠️ WARNING: Do not do this without first carefully reviewing the code of this PR to satisfy yourself it is safe.

curl -fsSL https://raw.githubusercontent.com/dotnet/maui/main/eng/scripts/get-maui-pr.sh | bash -s -- 36004

Or

• Run remotely in PowerShell:

iex "& { $(irm https://raw.githubusercontent.com/dotnet/maui/main/eng/scripts/get-maui-pr.ps1) } 36004"

[kubaflo]

Note

🤖 This review was automatically generated by a multi-model AI review system (Claude Opus 4.8, GPT-5.5, Gemini 3.1 Pro). Three models independently reviewed the code, then cross-pollinated their findings to produce this consolidated review.

Multi-Model Review — PR #36004 Round 1

Verdict: ✅ LGTM (Unanimous)
Confidence: High

Summary

Perfect unanimous approval. This PR correctly deflakes a pre-existing time-bomb in the release-readiness skill's test suite by replacing wall-clock-dependent assertions with deterministic synthetic fixtures.

What This PR Fixes

The Problem: Four live assertions hardcoded hasRecentActivity = $true against real branches (SR8, SR9, active-SR loop, preview6). These assertions were only true while the branch had a commit in the trailing 7 days. When SR8 went quiet for 7 days (normal end-of-cycle state), two assertions flipped red on 2026-06-18.

The Solution:

1. Removed wall-clock-dependent live assertions — Now asserts only that hasRecentActivity is a real [bool], never a date-dependent value
2. Added deterministic synthetic coverage — Throwaway temp git repo with commits at controlled dates (GIT_AUTHOR_DATE/GIT_COMMITTER_DATE at now−6d / now−8d / now−30d)
3. Tests real code — Calls the actual Get-RecentCommitCount helper (dot-sourced) and asserts exact counts for 7/10/60/1-day windows
4. Proper cleanup — try/finally cleans up both temp repo and mutated environment variables

Verification (All 3 Reviewers Agreed)

✅ Time-bombs removed — No more date-dependent assertions on live branches
✅ Deterministic fixture — Controlled commit dates, zero network, zero wall-clock dependence
✅ Complete coverage — All four recency windows (7/10/60/1-day) tested
✅ Proper cleanup — Temp repo and environment cleaned in finally block
✅ Suite results — Passed: 552, Failed: 0 (was 545/2 before with SR8 reds)
✅ Offline pass — New fixture assertions pass under -SkipE2E, confirming network-independence

Cross-Pollination Notes

• All 3 models: Independent LGTM with high confidence
• Opus: Verified fixture uses real Get-RecentCommitCount, proper cleanup, hermetic coverage
• GPT: Confirmed time-bomb removal and deterministic replacement
• Gemini: Validated synthetic fixture quality and safer E2E invariant
• Zero disagreements — Perfect alignment across all reviewers

This is a clean, low-risk deflaking fix that makes the test suite reliable and maintainable. Ready for merge.