Skip to content

Use shared MSTest core in UI test hosts - #12446

Open
Amaury Levé (Evangelink) wants to merge 10 commits into
mainfrom
dev/amauryleve/shared-core-ui-contract
Open

Amaury Levé (Evangelink) wants to merge 10 commits into
mainfrom
dev/amauryleve/shared-core-ui-contract

Conversation

@Evangelink

@Evangelink Amaury Levé (Evangelink) commented Nov 13, 2026 •

Copy link
Copy Markdown
Member

Pull request

Summary

Layer 1 of the replacement for #12435. Relates to #7225; builds on merged #12412 and #12428. #12435 and the native stack metadata remain unchanged.

Use the ordinary netstandard2.0/net462/net10.0/net11.0 cores for UI hosts, retaining the nonpackable core and existing Extensions-owned package layout. UI directory getters return null even with populated bags. Unsupported deployment fails active tests before their bodies; ignored folded/unfolded rows remain ignored.

Review fixes move deployment rejection after ignore decisions, declare forwarded UI APIs, and distinguish MSBuild wrapper exit 1 from direct MTP failed-test exit 2.

Two scoped CI-test repairs are included: fixture-owned retry connection grace, and explicit enabled-SDK environment ownership for three OpenTelemetry checks. The final tagged review fix explicitly owns the test's activity processor and checks provider shutdown before that local owner is cleaned up. Production retry/disposal policy and dependency versions are untouched; existing assertions, unexpected-delay guards, scheduling and serialization declarations remain intact.

Design and decision record

Shared-core UI contract, decisions and historical paired evidence.

Current published head: b97623a7069f8e78794370e83aa0185293a53728.
Coordinated frozen main ancestor: 94711f3b4db336920e628039fdd54437458ae441.

The94711 rebase preserves both upstream #12490 restore/coverage fixes and #12489 net462 timeout fix. The environment-isolation transfer 0e22cd431656c944906bdb547648d84369471c95 is based exactly on 500f4f0c50401dd1e1945a846b3d183a379d5ddf. The final processor-ownership commit is its direct child. Both change only OpenTelemetryProviderExtensionsTests.cs. All writes used exact observed-tip leases and independent ref/live-PR readbacks. No independent main chasing, merge, approval or stack metadata change.

Demonstration

No layout implementation changed. Shipping WinUI TRX assertions bind each identity to its result: six failed active deployment cases, one passed metadata-provider control, two ignored rows retaining disabled, direct host exit 2, and no executed-body marker.

Retry-fixture ordering/expiry/cancellation uses controlled completion sources, not wall-clock sleeps. OTel tests retain exporter shutdown/disposal assertions and arrange OTEL_SDK_DISABLED=false inside the existing restoring environment scope and matching writer lock.

Validation and regression proof

Release; Arcade SDK/templating 12.0.0-beta.26565.1, .NET SDK 12.0.100-alpha.1.26562.108, Visual Studio/MSBuild 18.13.12409.143. Extension TFMs: net10.0, net11.0, net462 and net472.

Final OTel class and tagged disposal review

Actual native Ubuntu CI reported metric exporter shutdown expected1/actual0. The unlocked SDK-building reader could race an existing writer of OTEL_SDK_DISABLED=true; the exact CI mid-build ordering was not captured. The repair scopes the three enabled-SDK checks under WithEnvironment(false) plus ResourceLock(WellKnownResources.EnvironmentVariables), preserving every assertion and existing DoNotParallelize.

The human-tagged processor-disposal thread is addressed by an explicit using SimpleActivityExportProcessor local passed to AddProcessor. Both resolved SDK versions guard repeated processor disposal. The added exporter.ShutdownCount == 1 assertion executes after the provider's scope and before local processor cleanup; the latter cannot mask missing provider shutdown. Existing lifecycle/exporter assertions, environment locks and both DoNotParallelize declarations are unchanged. The inline reply was read back (680 characters); the thread was not resolved on the reviewer's behalf. The latest live inventory shows all three review threads resolved externally and the PR approved.

The final one-file patch was tested on parent 0e22cd431656c944906bdb547648d84369471c95, then committed unchanged as b97623a7069f8e78794370e83aa0185293a53728. Patch SHA-256: 3B5121200CF9B828904CF95A751DD80D36E88BCCECBCE20FE2195A018F84D3A1; resulting test blob: 7d694a308e6a13dc6f4aeec393b9e5e4111fd403.

The extension-project Release build (.\build.cmd -c Release -msbuildEngine vs -projects test\UnitTests\Microsoft.Testing.Extensions.UnitTests\Microsoft.Testing.Extensions.UnitTests.csproj -bl) exited0 on all four TFMs with zero warnings. These exact runner commands executed from the repository root; $p was this session's local artifact directory, containing otel-32workers.runsettings. The retained settings contain <MSTest><Parallelize><Workers>32</Workers><Scope>MethodLevel</Scope></Parallelize></MSTest>. Local result/log paths below are retained evidence, not public artifact URLs.

foreach ($state in @('true','absent')) {
    if ($state -eq 'true') {
        $env:OTEL_SDK_DISABLED = 'true'
    } elseif (Test-Path Env:\OTEL_SDK_DISABLED) {
        Remove-Item Env:\OTEL_SDK_DISABLED
    }
    & .\.dotnet\dotnet.exe test --project test\UnitTests\Microsoft.Testing.Extensions.UnitTests\Microsoft.Testing.Extensions.UnitTests.csproj -c Release --no-build --filter 'FullyQualifiedName~OpenTelemetryProviderExtensionsTests.' --settings "$p\otel-32workers.runsettings" -p:UsingDotNetTest=true "-p:ArtifactsTestResultsDir=artifacts\TestResults\Release\otel-disposal-$state" "-bl:artifacts\log\Release\otel-disposal-$state.binlog" --show-test-results failed
}

Both invocations exited0: 172 executed/passed and zero skipped per state; 344 total,43 per TFM. All eight actual TRX files were independently parsed. Central OpenTelemetry1.20.1 was unchanged. The same selection on parent0e22 also passed344/344; processor ownership is preservation evidence, not a claimed pre-fix behavioral failure.

The native-equivalent target was also built and executed in this checkout, using the same final test bytes. With $env:TESTFX_BAZEL_OUTPUT_ROOT='Q:\tfb\n12446-dispose-642c9497', this command ran once with inherited OTEL_SDK_DISABLED=true and once with the variable removed:

.\eng\bazel.ps1 run //:extension_tests_app --jobs=2 --noshow_progress --show_result=0 '--' --filter 'FullyQualifiedName~OpenTelemetryProviderExtensionsTests' --minimum-expected-tests 1 --report-trx --report-trx-filename tests.trx --results-directory "Q:\tfb\n12446-dispose-642c9497\otel-disposal-$state"

Both invocations exited0: 43 executed/passed, zero skipped per state; 86 total,32 method-level workers. Both actual TRX files were parsed, rather than relying on minimum1 selection. The generated runfiles manifest resolves OpenTelemetry1.19.1, ProductVersion 1.19.1+5fbeba3a3d8bbd4f4235170ddeb6329fe0b8b86e, DLL SHA-256 6623430269ED62F1A8E097E3909B9C8D778AE72B9B69F096140C71D06E0EB043. Its net10 runtimeconfig uses Major roll-forward through pinned SDK 11.0.100-rc.3.26522.148. This is not a dependency bump or a consumer-package rebuild.

The independently inspected native-owner environment-isolation comparison remains separate: checkout 19be23d50e4f5be617ff4e1fb6acc97e4452ef5e plus only the one-file transfer patch, not an execution of the entire500f tree. Original test/provider blobs match500f. Transfer patch SHA-256 BEA2DC8F882B243FC633170989F8188CC7A75E8FBA5B427B2F96A23A930748CB: inherited-true old lifecycle case executed1/failed1/skipped0 at activityExporter.ShutdownCount expected1/actual0 (exit2); the fixed same case executed1/passed1/skipped0 (exit0). Both43-case class runs passed under true/absent. The CI failure was metric shutdown; this controlled inherited-state failure is not a captured reproduction of the precise CI mid-build interleaving.

Retry CI maintenance

Azure1683181 on earlier head3af reported RetryRecovery_BytesExactly16MiB/net10 and RetryRecovery_RootFlagOther/net11: Retry recovery must not schedule an unowned delay. A controlled probe reproduced the supported exit-first grace path against the overly strict fake. Same-source frozen-main Linux modules passed, so this was not dismissed as a proven baseline flake or hidden by a CI retry.

The repaired fixture owns only the supported cancellable one-second grace, observes/drains losing waits, and joins invocation cancellation with an independent cleanup bound. All recovery assertions and unexpected-delay guards remain unchanged.

.\.dotnet\dotnet.exe test --project test\UnitTests\Microsoft.Testing.Extensions.UnitTests\Microsoft.Testing.Extensions.UnitTests.csproj -c Release --no-build --filter "FullyQualifiedName~Microsoft.Testing.Extensions.UnitTests.Retry" -p:UsingDotNetTest=true -p:ArtifactsTestResultsDir=artifacts\TestResults\Release\retry-947-wave -bl:artifacts\log\Release\retry-947-wave.binlog --show-test-results failed

On the94711 tree immediately before the one-file OTel transfer: 1,432 independently discovered/executed/passed, zero skipped; exit0,358per TFM. The OTel transfer does not change those files. Normal method-level/CPU-count scheduling is preserved.

.\.dotnet\dotnet.exe test --project test\UnitTests\Microsoft.Testing.Extensions.UnitTests\Microsoft.Testing.Extensions.UnitTests.csproj -c Release --no-build --filter "FullyQualifiedName~RetryTests.RetryConnectionGracePeriod_ProtocolFixture" -p:UsingDotNetTest=true -p:ArtifactsTestResultsDir=artifacts\TestResults\Release\retry-fixed-new-cases -bl:artifacts\log\Release\retry-fixed-new-cases.binlog --show-test-results all

Five deterministic cases per TFM pass: 20 passed, zero skipped; exit0. The isolated actual old fake with identical tests yields 12 intended failures plus eight passed controls, zero skips; exit2. Both variants build. These cases are included in the1,432 rerun.

Earlier shipping and activation evidence

The new-Arcade37-project shipping/fixture pack and package verifier pass on all declared TFMs, zero warnings. Exact commands and scope are retained in the decision record and prior evidence.

Exact command Result
.\build.cmd -pack -c Release -msbuildEngine vs -projects artifacts\validation\pr12446-pack.slnf -bl Exit0; eight package-shape owners, generator and required fixture/dependency owners.
.\build.cmd -c Release -msbuildEngine vs -projects test\IntegrationTests\MSTest.Acceptance.IntegrationTests\MSTest.Acceptance.IntegrationTests.csproj -bl Exit0.
.\.dotnet\dotnet.exe test --project test\IntegrationTests\MSTest.Acceptance.IntegrationTests\MSTest.Acceptance.IntegrationTests.csproj -c Release --no-build --filter "FullyQualifiedName~WinUITests.DeploymentItemsFailBeforeBodiesIncludingInheritedAttributes" -p:UsingDotNetTest=true -p:ArtifactsTestResultsDir=artifacts\TestResults\Release\review-arcade-wave -bl:artifacts\log\Release\review-arcade-wave.binlog --show-test-results all Three planned/executed/passed, zero skipped; exit0; copied runtime hashes match the fresh package.

Shipping sources are unchanged. Earlier-head selections passed26acceptance and20context cases; these are not counted as newly rerun on final94711.

Azure1683108 app-model job executed 56 passed, zero failed/skipped on earlier head 984c2c635742d3d9699f488af36d25dff48e0f85. Published outer/child TRX and resolved packages verify ModernUwp_Net10, ModernUwp_Net11, and ModernUwp_UnsupportedDeployment: positives each have six passed results; the negative has six passed and two pre-body deployment failures. The outer negative passes MSBuild wrapper exit1. Local CodeCoverage launch failure remains historical setup context, not a missing contract on that revision. Earlier evidence is not final-head CI.

On final head b97623a7069f8e78794370e83aa0185293a53728, the Azure1683523 app-model job succeeded. Structured run46974020 independently confirms 56 executed/passed, zero skipped, including ModernUwp_Net10 (result100047), ModernUwp_Net11 (100052) and ModernUwp_UnsupportedDeployment (100054). The current-head outer wrapper-exit-1 assertion passes. Run46974022 additionally confirms 12/12 passed packaged WinUI/WinApp interoperability cases. These are final-head executions; detailed child-TRX/package-hash inspection above remains explicitly bound to earlier1683108 rather than being relabeled.

Current-head remote status

Azure1683523 completed/succeeded at16:37:59Z for source head b97623a7069f8e78794370e83aa0185293a53728, merge a389306fa9aad66dcecd37c597b2d843c03a506c. Raw merge parents are live-main ba9d282ba8cba9ac6afea36b1a9b2686aa6de883 and the exact published head; the branch retains its coordinated94711 ancestor.

The exact-head combined snapshot at16:40:03Z contains 21 SUCCESS,19 SKIPPED, zero failures and zero pending; no cancellation was excluded as superseded. Both native matrices are SUCCESS: Ubuntu, Windows. Azure aggregate, Linux Release, Windows Debug, Windows Release, both BuildXL configurations and app-model acceptance all succeeded. The19 skipped agentic-workflow stages are not counted as test execution proof. Earlier-head builds are not substituted for final-head CI; no checks, assertions or diagnostics were disabled.

Test-value evidence

Contract / exact cases Old behavior / evidence boundary Correct implementation
Retry exit-first grace, expiry and cancellation: RetryRecovery_ExitBeforeConnection, RetryRecovery_GraceExpires, RetryConnectionGracePeriod_ProtocolFixtureCancellationIsObserved Actual old fake with identical tests:12 intended failures/eight passed controls across four TFMs, zero skips; exit2. 20 focused cases pass; all included in1,432 passed/zero-skipped94711 rerun.
Enabled OTel SDK ownership: OpenTelemetryProvider_InvokesConfigurationDelegatesAndDisposesBothProviders and two SDK checks Native-owner1.19.1 controlled inherited-true old/fixed evidence: activity-exporter shutdown1/0 failure, then same case passes; exact CI race schedule unobserved. Final b976 test bytes: MSBuild1.20.1 class344 passed and own native1.19.1 class86 passed, zero skipped, true/absent inheritance and32workers.
Explicit processor ownership: EndToEnd_InstrumentationAndResource_ExportSpanWithSemanticConventionTagsAndResource Preservation: parent0e22 class344 passes. New pre-cleanup shutdown assertion statically predicts detection of missing provider shutdown; no disposal fault-injection run is claimed. Same class selection on final patch344 passes; own native selection86 passes. Existing shutdown assertions remain unchanged.
UI deployment ignores: Deployment_Reflection, Deployment_Rooting, Deployment_ReflectionFree Historical pinned-source comparison executes three intended old-guard failures; exact source/patch evidence is in the decision record. Same runtime/assertions pass all three fresh-package cases under Arcade26565.1; later infrastructure-only rebases are not new counterfactual executions.
Real UWP wrapper exit Earlier-head Azure1683108 child reports and resolved packages verify both positives and deployment negative, including wrapper exit1. Final-head Azure1683523 run46974020 confirms56/56 pass with the same three Modern UWP rows, including the outer wrapper-exit-1 assertion; run46974022 adds12/12 packaged WinUI/WinApp passes.

Retry repair patch SHA-256 on pre-rebase3af: 70987BE535748A4B5ED661233D0A5AD0CEB8C85094E76167B27D005DF1238A84; retained old-fake/support patch 6635A9068D117F6B1B6B4A5D6CB58723BF6198F95A5BB95B4AAB852BB0037A3B; identical test-file hash 70315183CB8F2D076D4CCC601B9C559CF2E57C45BC81697B40BFA7764472D4E9. Setup/compiler failures are not red proof. Reports/source patches/binlogs are retained; cleanup touched only named owned generated outputs, junctions and dumps, preserving source workspaces/shared caches. The final owned native Bazel server was shut down after validation; its named reproducible cache and reports were retained.

Keep Extensions as the existing package owner while unifying core builds and enforcing the UI runtime contract.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot AI balanced review requested due to automatic review settings November 13, 2026 10:02
@github-actions github-actions Bot added the state/needs-review Awaiting review from the team. label Nov 13, 2026

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🧪 Test quality review — PR #12446

Compared merge base 8dd78e1347417e400c3ef8d786059a8bb346162f with head 7b382c059ea63629db1f53d72632399af60c74ed. Evidence applies to this head, not every intermediate commit. This review is read-only and static; no tests were executed.

Test-value evidence

Contract Scenario / expected outcome Test and decisive assertion Counterfactual Status
TestContext directory properties return null in UWP/WinUI hosts, even when Properties has a same-named key Set Properties[dir]="not-a-ui-directory", expect null WinUITests/ModernUwpTests/ClassicUwpTests host-side Assert.IsNull(GetProperty(dir).GetValue(TestContext)) Removing the #if WINDOWS_UWP || WIN_UI null overrides makes the base GetProperty<string> return the seeded value, so the assertion fails Static prediction (killed)
[DeploymentItem] on a method or class (including inherited) fails the test before its body runs Expect Failed with the "not supported" message and no DEPLOYMENT_BODY_EXECUTED DeploymentItemsFailBeforeBodiesIncludingInheritedAttributes (3 source-gen modes) and ModernUwp_UnsupportedDeployment: the TRX identity:outcome sequence, plus Assert.Contains on the message and Assert.DoesNotContain on the marker Removing the TestMethodRunner guard makes the bodies run and Assert.Fail with the marker, so the outcome and message assertions fail Static prediction (killed)
DeploymentItemAttribute and the directory properties are now shared in modern core Core assembly exposes the type and properties; the UI extensions assembly does not CoreFrameworkAssemblyTests: Assert.IsNotNull(GetType(...)), Assert.IsNotNull(GetProperty(dir)), and a shared-core SHA-256 hash comparison Re-adding the #if exclusions makes the IsNotNull assertions fail Static prediction (killed)
WindowsApplicationModelPackageTests data rows flip DeploymentItem availability for modern UWP and WinUI Expect available for modern UWP/WinUI and unavailable for classic UWP Data-row expectations at lines 165-167 Reverting the shared-core change contradicts the rows Static prediction
TestContextTests: #if !WIN_UI guards removed Directory properties are null by default and use independent keys MissingAndNullPropertiesHaveDefinedDefaults and DirectoryPropertiesUseIndependentKeys Behavior-preserving on non-UI builds; expected to pass before and after Static prediction (green before and after)
Resources, xlf, csproj TFM simplification, docs, InternalAPI, LibraryProbe Resource string, build wiring and documentation None targeted. The csproj change is covered indirectly by the acceptance package tests. n/a Not applicable or unverified

No paired execution records for these revisions were available, so every row is a static prediction. CI green status was not treated as proof.

🧵 Parallel safety

PARALLEL-SAFETY RESULT
STATUS: PARTIAL

The specialist task was not run, so parallel-safety coverage is incomplete. This is not a clean result. By inspection only, the new acceptance tests generate a separate TestAsset per data row, and the shared TestContext.Properties writes are per-test instance state. Neither looked like a hazard, but this was not verified. Re-run with /parallel-audit for the full analysis.

Per-test assessment

Pseudo-mutation evidence below consists of static predictions, not executed mutation results. Production context for the #if WINDOWS_UWP/WIN_UI branches was read from the diff only; these branches are not buildable on the Linux runner, so conclusions are N/A / unverified for build behavior.

Result Quality Test Notes How to improve
Pass A (90–100) mod ModernUwpTests.ModernUwp_ConsumesUwpAssets_AndRunsPlainAndUiTestsThroughMtp The exact TRX identity:outcome sequence and message checks cover both rows. None
Pass A (90–100) new WinUITests.DeploymentItemsFailBeforeBodiesIncludingInheritedAttributes Asserts exact identities, outcomes, message and no body execution in all 3 modes. None
Pass B (80–89) mod CoreFrameworkAssemblyTests (asset probe tests) The probe asserts type and property presence and a shared-core hash. None
Pass B (80–89) mod WindowsApplicationModelPackageTests (DeploymentItem data rows) Data rows encode the new availability matrix. None
Pass B (80–89) mod WinUITests / ClassicUwpTests directory null assertions Seeds Properties and asserts null through reflection for all 6 properties. None
Pass B (80–89) mod TestContextTests.MissingAndNullPropertiesHaveDefinedDefaults Exact null assertions for the six directory properties. None
Pass B (80–89) mod TestContextTests.DirectoryPropertiesUseIndependentKeys Distinct values asserted per property. None

No inline findings met the publication bar, and no existing-thread duplicates were considered. Existing review threads were not read in this run.

<sub>This advisory review is heuristic and non-blocking.

Re-run the combined review with /review-tests, or only the parallel-safety specialist with /parallel-audit.</sub>

🤖 Automated content by GitHub Copilot. Generated by the Test Quality Review on PR (on open / sync) workflow. · copilot · auto · ~33 AIC · ⌖ ~1 AIC · ⊞ 8.1K

@microsoft-github-policy-service microsoft-github-policy-service Bot added state/needs-review Awaiting review from the team. and removed state/needs-review Awaiting review from the team. labels Nov 13, 2026
@github-actions github-actions Bot added state/needs-review Awaiting review from the team. and removed state/needs-review Awaiting review from the team. labels Nov 13, 2026

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Note

🤖 Automated review by GitHub Copilot. Generated by the Expert Code Review workflow. To request a follow-up action, reply by tagging @copilot directly.

Scope: shared-core UI contract, layer 1. It removes the UWP/WinUI TFMs from TestFramework, makes the directory getters on UI TestContext return null, and makes the adapter fail tests that use DeploymentItemAttribute. I reviewed the diff only, without building or running anything.

Finding (1 inline):

  • The Modern UWP deployment test expects exit code 1, but MTP uses 2 (AtLeastOneTestFailed) when tests fail. See the inline comment.

Observations (no action required, not proven defects):

  • The TestTempDirectory override in TestContextImplementation.TempDirectory.cs is guarded #if !WINDOWS_UWP && !WIN_UI, so on UI hosts it falls back to the base null. This matches the stated contract.
  • The five new override ... .get -> string? lines in InternalAPI.Unshipped.txt are for internal API, so there is no public API surface change. TestContext now exposes these members for UWP/WinUI TFMs, but those TFMs are consumed from the netstandard2.0 and net10.0 core, whose PublicAPI files already declare them.
  • DeploymentItemNotSupportedForUi was added to the resx and the xlf files, and the {Locked=...} tokens occur verbatim in the value.
  • Evidence gaps (documented in the decision record): real UWP activation and a full Bazel run were not executed, and the shared-core hash-identity cases have no counterfactual (mutation) run. The ModernUwp_UnsupportedDeployment row is therefore unverified, which is how the exit-code issue could slip through.

Test-value evidence: the PR record cites a red/green comparison (13 passed / 4 failed without the change, 17 passed with it) at revision e22e899f. I did not re-run it.

@microsoft-github-policy-service microsoft-github-policy-service Bot added state/needs-review Awaiting review from the team. and removed state/needs-review Awaiting review from the team. labels Nov 13, 2026
@github-actions github-actions Bot added state/needs-review Awaiting review from the team. and removed state/needs-review Awaiting review from the team. labels Nov 13, 2026

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

The deployment guard runs before data-source and row ignore checks, causing ignored data-driven UWP/WinUI tests with DeploymentItem to fail instead of being skipped.

1 open issue
What changed in this PR

Prepares MSTest for the UWP/WinUI package separation in #7225 by sharing ordinary core assemblies across UI hosts, while retaining Extensions as the package owner and preserving existing package paths.

Changes:

  • Removes UI-specific core builds and exposes shared directory/deployment APIs.
  • Makes UI-host directory getters return null and rejects unsupported deployment attributes before test bodies run.
  • Updates consumer tests, API expectations, localized resources, and the decision record.
Confidence at a glance — 🟡 Static review; execution evidence not independently verified

Test-value evidence: For runtime revision e22e899febaeee44ea3af385403e1c65c18164bb, the decision record reports four intended failures with the behavior removed, followed by 17 passing cases after restoration. These runs were not independently verified here; real UWP activation and full native Bazel execution remain documented gaps.

File Description
test/​UnitTests/​TestFramework.UnitTests/​Extensions/​TestContextTests.cs Enables directory-contract checks on WinUI builds.
test/​IntegrationTests/​MSTest.Acceptance.IntegrationTests/​WinUITests.cs Tests null directories and deployment rejection across metadata modes.
test/​IntegrationTests/​MSTest.Acceptance.IntegrationTests/​WindowsApplicationModelPackageTests.cs Updates packaged UI API expectations.
test/​IntegrationTests/​MSTest.Acceptance.IntegrationTests/​ModernUwpTests.cs Adds deployment-failure and null-directory assertions.
test/​IntegrationTests/​MSTest.Acceptance.IntegrationTests/​CoreFrameworkAssemblyTests.cs Checks shared-core hashes and API availability.
test/​IntegrationTests/​MSTest.Acceptance.IntegrationTests/​ClassicUwpTests.cs Checks null directory getters in classic UWP.
src/​TestFramework/​TestFramework/​TestFramework.csproj Restricts core builds to ordinary target frameworks.
src/​TestFramework/​TestFramework/​TestContext.cs Exposes directory properties and documents UI behavior.
src/​TestFramework/​TestFramework/​Attributes/​DeploymentItemAttribute.cs Exposes the attribute on modern UI builds and documents rejection.
src/​Adapter/​MSTestAdapter.PlatformServices/​Services/​TestContextImplementation.Properties.cs Overrides all six UI directory getters to return null.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.zh-Hant.xlf Adds the deployment diagnostic to the Traditional Chinese catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.zh-Hans.xlf Adds the deployment diagnostic to the Simplified Chinese catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.tr.xlf Adds the deployment diagnostic to the Turkish catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.ru.xlf Adds the deployment diagnostic to the Russian catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.pt-BR.xlf Adds the deployment diagnostic to the Brazilian Portuguese catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.pl.xlf Adds the deployment diagnostic to the Polish catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.ko.xlf Adds the deployment diagnostic to the Korean catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.ja.xlf Adds the deployment diagnostic to the Japanese catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.it.xlf Adds the deployment diagnostic to the Italian catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.fr.xlf Adds the deployment diagnostic to the French catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.es.xlf Adds the deployment diagnostic to the Spanish catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.de.xlf Adds the deployment diagnostic to the German catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​xlf/​Resource.cs.xlf Adds the deployment diagnostic to the Czech catalog.
src/​Adapter/​MSTestAdapter.PlatformServices/​Resources/​Resource.resx Defines the unsupported-deployment message and localization markers.
src/​Adapter/​MSTestAdapter.PlatformServices/​InternalAPI/​InternalAPI.Unshipped.txt Records new directory-property overrides.
src/​Adapter/​MSTestAdapter.PlatformServices/​Execution/​TestMethodRunner.cs Adds pre-body deployment rejection for modern UI hosts.
eng/​bazel/​windows/​LibraryProbe.cs Updates native probe expectations for shared APIs.
docs/​decisions/​shared-core-ui-contract.md Records the staging contract, rationale, validation, and limitations.

Review effort: Balanced

@microsoft-github-policy-service microsoft-github-policy-service Bot added state/needs-review Awaiting review from the team. and removed state/needs-review Awaiting review from the team. labels Nov 13, 2026
@github-actions github-actions Bot added state/needs-review Awaiting review from the team. and removed state/needs-review Awaiting review from the team. labels Nov 13, 2026
@microsoft-github-policy-service microsoft-github-policy-service Bot added the state/needs-review Awaiting review from the team. label Nov 16, 2026
@github-actions github-actions Bot added state/approved Proposal approved; ready for implementation. and removed state/needs-review Awaiting review from the team. state/approved Proposal approved; ready for implementation. labels Nov 16, 2026
@microsoft-github-policy-service microsoft-github-policy-service Bot added the state/needs-review Awaiting review from the team. label Nov 16, 2026
@github-actions github-actions Bot added state/approved Proposal approved; ready for implementation. and removed state/needs-review Awaiting review from the team. state/approved Proposal approved; ready for implementation. labels Nov 16, 2026

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Pending final-head Windows app-model CI and unavailable retained artifacts prevent independent validation of UWP/WinUI activation and package provenance, requiring maintainer confirmation.

No open issues 🎉

1 resolved since last review

Review effort: Balanced

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot AI balanced review requested due to automatic review settings November 16, 2026 14:47
@github-actions github-actions Bot added state/approved Proposal approved; ready for implementation. and removed state/approved Proposal approved; ready for implementation. labels Nov 16, 2026
@microsoft-github-policy-service microsoft-github-policy-service Bot added the state/needs-review Awaiting review from the team. label Nov 16, 2026
@github-actions github-actions Bot added state/approved Proposal approved; ready for implementation. and removed state/needs-review Awaiting review from the team. state/approved Proposal approved; ready for implementation. labels Nov 16, 2026

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Shared-core changes affect UWP and modern UI runtime/package binding, but final-head CI is pending and Windows activation and matching artifacts still require maintainer verification.

No open issues 🎉

Review effort: Balanced

Preserve the rebased branch and subsequent review fixes without rewriting local history. The resulting tree matches the fetched remote head.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Preserve both the owned retry connection-grace lifecycle and upstream connected-child callbacks when resolving the fixture conflict.

Validation: pinned Release builds and package verification passed; 1,204 focused retry, telemetry, CLI, terminal, and packaged-consumer cases passed with no skips.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot AI balanced review requested due to automatic review settings November 16, 2026 20:01
@github-actions github-actions Bot added state/approved Proposal approved; ready for implementation. and removed state/approved Proposal approved; ready for implementation. labels Nov 16, 2026

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟢 Approval recommended

No active review issues block approval of the shared-core UI contract changes.

No open issues 🎉

Review effort: Balanced

@github-actions

Copy link
Copy Markdown
Contributor

Final test-triage resolution — Azure Pipelines build 1684105

Result: no actionable test failure, but inconclusive because the evidence is incomplete.

The build microsoft.testfx succeeded. The collector read about 112.5k results: 111,669 passed, 894 skipped and 8 failed. It also reported evidenceIncomplete: true and one artifact download failure. The timeline had no failed or warned records, and there were no crash or hang diagnostics. Because of the gaps, I can't call the tests clean.

The 8 failed records all come from the Windows_App_Model_Diagnostics_Attempt1 artifact, and each looks like an intentional probe:

  • WinUIMtpPackagedApp.UnitTest1.RetryFailsFirstAttempt and PackagedWinUI...RetryTestMethod_PassesOnSecondAppContainerHost are retry probes. They deliberately fail the first attempt (or host).
  • NativeTests.Outcome (4 records) fails with "expected host failure" in NativeSidecar report-path scenarios. That is the expected outcome of those scenarios.
  • ModernUwpee5a6a6f9016.InheritedMethodDeployment and InheritedClassDeployment fail with "DeploymentItemAttribute is not supported by UWP or WinUI test hosts". That is the expected diagnostic for those test cases.

These are probes output rather than product regressions. I found no proven fail-then-pass flakiness: the retry records are TRX probe artifacts, not CTRF retry metadata. The slowness checks found 0 regressions. I did not create an issue because nothing meets the durable-issue thresholds.

Confidence: medium. The conclusion rests on the artifact paths and failure messages, plus the successful build result.

🤖 Automated content by GitHub Copilot. Generated by the Pipeline Test Triage workflow. · copilot · auto · ~28 AIC · ⌖ ~1 AIC · ⊞ 10.1K

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

state/approved Proposal approved; ready for implementation.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants