Backport genuinely-missing main commits to rel/4.3

[Evangelink]

Backports the genuinely-missing (non-vstest-removal) commits from main to rel/4.3. Every commit was cherry-picked with -x; the full solution builds clean (build.cmd -c Debug: 0 warnings, 0 errors).

Backported (8)

Perf

• perf: skip TCS bridge in ExecuteTestAsync when no ExecutionContext is captured (perf: skip TCS bridge in ExecuteTestAsync when no ExecutionContext is captured #9636)
• perf: avoid GetParameters() allocation in test-invocation hot path (perf: avoid GetParameters() allocation in test-invocation hot path #9671)

Tests

• test: add edge case tests for MSTEST0045/0050/0060 analyzers (test: add edge case tests for MSTEST0045/0050/0060 analyzers #9615)
• test: add edge case tests for MSTEST0020/0021 analyzers (test: add edge case tests for MSTEST0020/0021 analyzers #9669)
• Re-enable LocalizationTests and LocalizationFailingTests (Re-enable LocalizationTests and LocalizationFailingTests #9643)

Docs / infra

• docs: fix Changelog-Platform links and add Emit ::warning annotations for skipped tests in GitHubActionsReport #9641 entry (docs: fix Changelog-Platform links and add #9641 entry #9667) — resolved conflict to add only the Emit ::warning annotations for skipped tests in GitHubActionsReport #9641 entry (already backported); dropped the main-only Add --report-azdo-groups/--report-azdo-annotations on|off toggles #9542/Re-print errored assemblies in dotnet test end-of-run recap #9545/Add server-initiated session cancellation to the dotnet test IPC protocol (#8691) #9549 entries
• Add scheduled workflow to bump bundled Playwright (fixes Update bundled Playwright for .NET version in MSTest SDK #9362) (Add scheduled workflow to bump bundled Playwright (fixes #9362) #9666)
• Document GH_AW_GITHUB_TOKEN removal and enterprise 8-day PAT policy (Document GH_AW_GITHUB_TOKEN removal and enterprise 8-day PAT policy #9588)

Intentionally excluded

• vstest removal / platform-agnostic decoupling (out of scope): Abstract VSTest discovery sink in PlatformServices (Phase 3 of platform-agnostic effort) #9625, Relocate the UnitTestElement<->TestCase conversion to the adapter (Phase 6e-3b) #9626, Drop the ObjectModel package reference from PlatformServices (Phase 7 capstone) #9633
• Already backported to rel/4.3: Transfer ownership in ReflectionMetadataHook.Register instead of cloning #9586 (as Transfer ownership in ReflectionMetadataHook.Register instead of cloning by @Evangelink in #9586 (backport to rel/4.3) #9589), perf: reduce allocations in data-driven test execution hot path #9617 (as perf: reduce allocations in data-driven test execution hot path by @Evangelink in #9617 (backport to rel/4.3) #9619), perf: skip intermediate Dictionary alloc in CloneForDataDrivenIteration #9637 (as perf: skip intermediate Dictionary alloc in CloneForDataDrivenIteration by @Evangelink in #9637 (backport to rel/4.3) #9639); perf: cache GetParameters() and ReflectionTestMethodInfo across data rows #9614 (functional change already present via perf: reduce allocations in data-driven test execution hot path by @Evangelink in #9617 (backport to rel/4.3) #9619)
• Depends on main-only infra: Avoid double TestCase construction when filtering during discovery (#9568) #9665 uses GetOrCreateHostTestCase(), introduced by the un-backported vstest decoupling (Relocate the UnitTestElement<->TestCase conversion to the adapter (Phase 6e-3b) #9626)
• Would misrepresent rel/4.3: docs: add missing changelog entries for #9542, #9545, #9549, #9581 #9613 documents main-only features (Add --report-azdo-groups/--report-azdo-annotations on|off toggles #9542/Re-print errored assemblies in dotnet test end-of-run recap #9545/Add server-initiated session cancellation to the dotnet test IPC protocol (#8691) #9549/Fix CloneWithUpdatedSource mutating this instead of the clone #9581) that never shipped in rel/4.3
• Auto-generated noise: [main] Update dependencies bumps and OneLocBuild check-ins

[github-actions]

🧪 Test quality grade — PR #9692

Grade	Test	Notes
B (80–89)	new DuplicateTestMethodAttributeAnalyzerTests. WhenDerivedAttributeIsFirstAndTestMethodIsSecond_ CodeFix_ KeepsDerivedAttribute	Body is ~52 lines due to embedded code strings; consider extracting the custom attribute boilerplate into a shared helper to reduce verbosity.
B (80–89)	new PreferConstructorOverTestInitializeAnalyzerTests. WhenTestClassHasOnlyParameterizedCtorAndTestInitialize_ CodeFix_ MergesIntoParameterizedCtor	Complex merge scenario is well-explained by the comment; body over 30 lines — trimming the embedded field declarations slightly would bring this to A.
B (80–89)	new PreferConstructorOverTestInitializeAnalyzerTests. WhenTestInitializeMethodInNonTestClass_ Diagnostic	Good edge case with informative comment explaining both diagnostic and fix; body just over 30 lines — a minor trim would reach A.
B (80–89)	new PreferDisposeOverTestCleanupAnalyzerTests. WhenTestCleanupMethodInNonTestClass_ Diagnostic	IDisposable conversion fix well-covered with comment; body is 32 lines — trimming the embedded code slightly would suffice.
B (80–89)	new UseCooperativeCancellationForTimeoutAnalyzerTests. WhenTimeoutAttributeOnAsyncTestMethod_ Diagnostic	Async-method edge case clearly verified with both diagnostic location and fix output; body 36 lines — shortening the test method stub would improve conciseness.
B (80–89)	new UseCooperativeCancellationForTimeoutAnalyzerTests. WhenTimeoutAttributeWithNamedArgument_ Diagnostic	Named-argument form well-covered; body just over 30 lines — a minor trim would be enough to reach A.
A (90–100)	new DuplicateTestMethodAttributeAnalyzerTests. WhenTestMethodHasDuplicateAttributesOutsideTestClass_ Diagnostic	Well-commented edge case; diagnostic location precisely marked via [|...|], clear single-behavior focus in ~20 lines.
A (90–100)	new GlobalTestFixtureShouldBeValidAnalyzerTests. WhenGlobalTestCleanupWithDerivedTestClassAttribute_ NoDiagnostic	Negative test verifying attribute inheritance is respected; comment references the analyzer's Inherits(testClassAttributeSymbol) guard directly.
A (90–100)	new GlobalTestFixtureShouldBeValidAnalyzerTests. WhenGlobalTestInitializeInGenericClass_ Diagnostic	Comment links the test to the allowGenericType: false parameter; concise, precisely scoped edge case in ~18 lines.
A (90–100)	mod LocalizationFailingTests. Execution_ WithFailingTest_ OutputContainsTranslatedFailureSummary	Multiple assertions: exit code + normalized French string checks; data-driven per TFM. Re-enabled from class-level ignore.
A (90–100)	mod LocalizationTests. Execution_ WithFrenchLocale_ OutputContainsTranslatedSummary	Thorough: exit code + 5 positive + 2 negative assertions covering all translated count labels. Re-enabled from class-level ignore.
A (90–100)	mod LocalizationTests. Execution_ WithSpanishLocale_ OutputContainsTranslatedSummary	Good parallel structure with the French test; direct AssertOutputContains (no normalization) is appropriate for Spanish. Re-enabled from class-level ignore.
A (90–100)	mod LocalizationTests. Execution_ WithTestingPlatformUILanguage_ TakesPrecedenceOverDotnetCLI	Clean precedence test with comment explaining expected outcome; exit code + positive and negative assertions. Re-enabled from class-level ignore.
A (90–100)	new UseCooperativeCancellationForTimeoutAnalyzerTests. WhenTimeoutAttributeOnMethodInNonTestClass_ Diagnostic	Non-test-class edge case cleanly covered; both diagnostic and fix verified in ~28 lines — the most concise of the three new timeout tests.

_{This advisory comment was generated automatically. Grades are heuristic
and informational — they do not block merging. Re-run with
/grade-tests.}

🤖 Automated content by GitHub Copilot. Posted via a maintainer's GitHub token, so it appears under their account — the account owner did not write or approve this content personally. Generated by the Grade Tests on PR (on open / sync) workflow. · 149.6 AIC · ⌖ 9.3 AIC · ⊞ 9.5K · [◷]( · ◷)

[github-actions]

Note

🤖 Automated review by GitHub Copilot. Posted via a maintainer's GitHub token, so it appears under their account — the account owner did not write or approve this content personally. Generated by the Expert Code Review workflow. To request a follow-up action, reply by tagging @copilot directly.

Expert Review — PR #9692

Dimension	Severity	Verdict	Notes
1. Algorithmic Correctness	MAJOR	✅	GetInvokeResultWithParametersAsync correctly threads the cached ParameterInfo[] through all call sites. ConstructGenericMethod signature and internal usage updated consistently.
2. Threading & Concurrency	BLOCKING	✅	TestMethodInfo.ParameterTypes uses benign lazy-init (field ??=); worst case is double-compute of an equivalent array. No shared-mutable-state concerns.
3. Security & IPC Contract Safety	BLOCKING	✅	No new deserialization or IPC changes. Workflow uses contents: read at top level with write scoped to the update job only.
4. Public API & Binary Compatibility	BLOCKING	✅	All new/changed methods are internal. No public surface changes.
5. Performance & Allocations	MAJOR	✅	Core goal of the PR — GetParameters() allocation eliminated from the hot path by caching via ParameterTypes property. Clean split between hot-path (GetInvokeResultWithParametersAsync) and non-hot (GetInvokeResultAsync wrapper).
6. Cross-TFM Compatibility	MAJOR	✅	Changes compile across all TFMs (net462/netstandard2.0/net8.0/net9.0). No TFM-specific APIs introduced.
7. Test Coverage & Quality	MAJOR	✅	New analyzer edge-case tests follow existing patterns (VerifyAnalyzerAsync/VerifyCodeFixAsync). Localization tests re-enabled.
8. Error Handling & Diagnostics	MAJOR	✅	RuntimeError raised with context in the Python script. C# error paths unchanged (same TestFailedException messages).
9. Naming & Style	MINOR	✅	arguments == null → arguments is null ✓. methodParameters?.Length ?? 0 → methodParameters.Length (now non-nullable) ✓.
10. Documentation & Comments	MINOR	✅	Good inline comments explaining the thin-wrapper pattern in MethodInfoExtensions. README updated with 8-day PAT policy and fast-unblock instructions.
11. Resource & Lifecycle Management	MAJOR	✅	N/A — no IDisposable or CancellationToken flow changes.
12. Localization	MAJOR	✅	N/A — no new user-facing strings.
13. Build & Project File Quality	MAJOR	✅	N/A — no .csproj / .props / .targets changes.
14. Backward / Forward Compatibility	MAJOR	✅	N/A — no wire-format changes.
15. Dependency Hygiene	MAJOR	✅	N/A — no new package references.
16. Changelog / Release Notes	MINOR	✅	#9641 entry added. Old testanywhere links corrected to testfx.
17. Agentic Workflow Quality	MAJOR	✅	N/A — no agentic workflow .md sources changed.
18. TODO Comment Policy	MINOR	✅	No bare TODO comments introduced.
19. CI/CD & Workflow Correctness	MAJOR	✅	Actions pinned to full SHAs (actions/checkout@34e11..., actions/setup-python@a26af..., peter-evans/create-pull-request@67ccf...). Least-privilege permissions. Concurrency control prevents racing updates.
20. Inline Test Infrastructure	MAJOR	✅	N/A — no acceptance test duration assertions.
21. MSBuild / SDK Quality	MAJOR	✅	N/A — no props/targets changes.
22. Python / Shell Script Quality	MAJOR	⚠️	Minor: _stable_tuple(current) can return None causing TypeError if the props file ever holds a prerelease string. Low practical risk. See inline comment.

Overall Verdict: LGTM

Well-crafted backport. The perf optimization in MethodInfoExtensions is the highest-impact change — it correctly eliminates per-invocation GetParameters() allocations on the data-driven test hot path by threading a lazily-cached ParameterInfo[] from TestMethodInfo.ParameterTypes. The new Playwright bump workflow follows all CI security best practices. Test additions are thorough edge-case coverage for existing analyzers.