Dogfood dotnet test for Mono.Android device tests#11224
Open
jonathanpeppers wants to merge 64 commits into
Open
Dogfood dotnet test for Mono.Android device tests#11224jonathanpeppers wants to merge 64 commits into
dotnet test for Mono.Android device tests#11224jonathanpeppers wants to merge 64 commits into
Conversation
jonathanpeppers
force-pushed
the
jonathanpeppers/dogfood-dotnet-test
branch
from
31e7d49 to
7b8a944
Compare
Member
Author
|
With MTP, test filtering works via command-line args passed to I've now added this support to
The old
Everything gets translated to a |
jonathanpeppers
force-pushed
the
jonathanpeppers/dogfood-dotnet-test
branch
from
04b483b to
8bc3984
Compare
…form (MTP) - Revamp `TestRunner.Core/TestInstrumentation.cs` to use MTP + NUnit (following the androidtest template pattern with `AddNUnit()`) - Delete `TestRunner.NUnit/`, `TestRunner.xUnit/`, `Xamarin.Android.NUnitLite/`, `src-ThirdParty/NUnitLite/` - Update `Mono.Android.NET-Tests` and `JcwGen-Tests` to use new `TestInstrumentation` base class with stock NUnit 4.3.2 - Update CI YAML to use `dotnet test --no-build` instead of `-t:RunTestApp` - Update `NUnitReferences.projitems` to NUnit 4.3.2 + adapter 5.0.0 - Clean up `Directory.Build.props`, installer targets, and solution files Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The NUnit 4.x upgrade broke all host-side test projects because NUnit 4 removed classic assert APIs (Assert.IsTrue, FileAssert, etc.). Only NUnit3TestAdapter 5.0.0 is needed (for AddNUnit() MTP support) in the device test projects. NUnit itself stays at 3.13.3, and the shared NUnitReferences.projitems is restored to its original versions. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The publish step was skipped because testResultsFiles was empty. Add --logger trx and --results-directory to produce TRX files, and use a deterministic file path for PublishTestResults@2. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Restore XA4313 AndroidError for Xamarin.Android.NUnitLite in Common.targets - Restore NUnitLite in XA4313 test Values in BuildTest.cs - Restore GenerateAssemblyInfo=false and AssemblyInfo.cs in TestRunner.Core - Delete unused MainActivities and Main.axml layout - Move MyFragment to its own file (still needed by FragmentFixup tests) - Fix TRX publishing: copy device-pulled TRX to expected path instead of relying on --logger trx which doesn't work with Microsoft.Android.Run Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Root cause: dotnet test uses VSTest mode by default, which runs vstest.console to discover tests from DLLs. Android device tests need MTP mode so dotnet test uses RunCommand/RunArguments to invoke Microsoft.Android.Run, which runs am instrument on the device. Changes: - Add global.json with MTP runner to each device test project directory (matches the androidtest template pattern) - Add IsTestingPlatformApplication=true to device test csprojs so the MTP mode of dotnet test recognizes them - Add GenerateTestingPlatformEntryPoint=false since Android projects have their own entry point - Update apk-instrumentation.yaml to cd into the project directory before running dotnet test, so it finds the project-local global.json - Fix NUnitInstrumentation.cs -> TestInstrumentation.cs in Compile items - Remove stale TrimmerRootAssembly refs to deleted projects - Add test exclusion support via ExcludedCategories/ExcludedTestNames Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
ComputeRunArguments needs intermediate build outputs (AndroidManifest.xml) to resolve the package name, and the app must be installed on the device before Microsoft.Android.Run can invoke am instrument. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This project previously got NUnit transitively via Xamarin.Android.NUnitLite. Now that NUnitLite is deleted, it needs a direct NUnit reference. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Define NUnitVersion in Configuration.props and reference it everywhere instead of copying 3.13.3 into each csproj. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
TestRunner.Core.NET.csproj did not import Configuration.props, so $(NUnitVersion) was empty causing NU1015. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
NUnit3TestAdapter targets netcoreapp3.1/net462 which are not compatible with Android TFMs, so AddNUnit() and MTP types are unavailable on device. Instead, use NUnitTestAssemblyRunner to run tests directly and generate TRX output that Microsoft.Android.Run pulls via adb and reports through MTP on the host side. Also add NUnit compatibility shim for TestFixtureSetUpAttribute and TestFixtureTearDownAttribute which were removed from stock NUnit but are used by Java.Interop submodule test code. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Previous edit accidentally removed the <ItemGroup> opening tag around ProjectReferences when removing NUnit3TestAdapter, leaving orphaned closing tags that caused Step_GenerateCGManifest XML parse failure. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
NUnit 3.13.3 filter classes (CategoryFilter, NotFilter, etc.) are internal. Use TestFilter.FromXml() with NUnit XML filter syntax instead. Remove [Obsolete] from TestFixtureSetUp/TearDown compatibility shims since warnings-as-errors treats them as build errors. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The Mono.Android.NET-Tests project does not have ImplicitUsings enabled, so explicit using directives are needed for System, System.Collections.Generic, and Android.OS. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
NUnit 3.13.3 DefaultTestAssemblyBuilder.Build(Assembly, ...) calls AssemblyHelper.GetAssemblyPath() which accesses Assembly.CodeBase. On .NET Android, assemblies are in a single-file bundle so CodeBase throws NotSupportedException. Fix by using the string-based Load(string, settings) overload with Assembly.FullName. NUnit resolves this via Assembly.Load(AssemblyName) which finds the already-loaded assembly without needing CodeBase. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
jonathanpeppers
force-pushed
the
jonathanpeppers/dogfood-dotnet-test
branch
from
8bc3984 to
cec4ac5
Compare
The previous fix (using assembly.FullName with the string-based Load overload) silently failed — NUnit sets AssemblyName.Name to the full display name string, which does not resolve to the loaded assembly. Replace DefaultTestAssemblyBuilder with a custom AndroidTestAssemblyBuilder that redirects Build(Assembly, ...) through the string overload using assembly.GetName().Name (simple name). This avoids Assembly.CodeBase entirely while correctly resolving the already-loaded assembly via Assembly.Load(AssemblyName). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add --report-trx and --results-directory to dotnet test so MTP generates a TRX file on the host side. This replaces the fragile find+copy step that searched $TMPDIR/AndroidTestResults for a device-pulled TRX file (which was not being found). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…publishing NUnit 3.13.3 (from NuGet) enforces that TestCaseSource must refer to static fields. The old bundled NUnitLite allowed instance fields. Make EncodingTestData, NinePatchDrawables, and AssetPngs static. Restore ExcludedTestNames for TrimmableTypeMap mode that were in the old NUnitInstrumentation but missing from the new TestInstrumentation subclass. Add --report-trx to dotnet test so MTP generates a TRX file on the host side for PublishTestResults@2. Remove the broken find+copy step that was searching $TMPDIR/AndroidTestResults and not finding anything. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The old TestApks.targets passed ExcludeCategories via am instrument -e extras. AndroidTestAdapter does not support extras, so bake the category exclusions into the TestInstrumentation subclass using RuntimeFeature checks instead. This excludes CoreCLRIgnore/NTLM on CoreCLR and NativeTypeMap/Export on trimmable typemap, matching the old behavior. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…od-dotnet-test # Conflicts: # THIRD-PARTY-NOTICES.TXT # tests/Mono.Android-Tests/Mono.Android-Tests/Xamarin.Android.RuntimeTests/NUnitInstrumentation.cs
The --report-trx flag is forwarded by MTP to the test host process (Microsoft.Android.Run). Without registering the TrxReport extension, the test host rejects the flag as unknown. Add the Microsoft.Testing.Extensions.TrxReport package and call AddTrxReportProvider() so MTP can generate TRX files on the host side. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
|
Azure Pipelines successfully started running 1 pipeline(s). |
simonrozsival
pushed a commit
that referenced
this pull request
May 29, 2026
…#11523) * Fix UrlEscaping_Bug43411 to actually run and assert correctly Two issues with this test on main: 1. The method was declared `void UrlEscaping_Bug43411 ()` (non-public). NUnitLite silently skipped non-public `[Test]` methods, so the test has been effectively dead for years. Discovered while migrating to stock NUnit 3 via `dotnet test` in PR #11224, where NUnit 3 surfaces the method and it fails. 2. The assertion compared the request URL against `Uri.ToString ()`, which returns the human-readable form and unescapes safe characters like `%20` to a literal space. The canonical, encoded form is `Uri.AbsoluteUri` — switch to that so percent-encoding is preserved. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Address review: make test async and verify URL on the listener side - Make `UrlEscaping_Bug43411` (and the `UrlEscaping_TestUrl` helper) `async Task` and `await` `SendAsync` instead of blocking with `.Wait ()`. - The previous listener callback unconditionally set `failed = true`, so `Assert.IsNull (failed)` always failed (`Expected: null, But was: True`). Capture `HttpListenerContext.Request.RawUrl` on the server side instead and compare it against the expected `Uri.PathAndQuery` — this is the verification the test was clearly meant to do (confirm that percent-encoded characters survive the round trip to the server). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…od-dotnet-test # Conflicts: # tests/Mono.Android-Tests/Mono.Android-Tests/Xamarin.Android.Net/AndroidMessageHandlerIntegrationTests.cs
Reverted the using directive reorder, the Console -> Android.Util.Log
conversion in ThreadReuse, and the try/catch in DoNotLeakWeakReferences.
Only keep what is strictly required for the migration: removing the two
`Java.Lang.JavaSystem.LoadLibrary ("reuse-threads")` calls from
RegisterTypeOnNewNativeThread and ThreadReuse, since the library is now
loaded once in TestInstrumentation.OnCreate on the main thread.
If Console.WriteLine on the native callback thread or the unhandled
exception in DoNotLeakWeakReferences resurface as "0 tests ran" in CI,
they will be addressed separately with logcat evidence in hand.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
jonathanpeppers
changed the title
dotnet test for device testsdotnet test for Mono.Android device tests
Member
Author
|
/review |
|
✅ Android PR Reviewer completed successfully! |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
🤖 Code Review Summary
Verdict:
Overview
Solid migration from NUnitLite to dotnet test + MTP. The new TestInstrumentation base class is well-structured with excellent comments (especially the NumberOfTestWorkers=0 rationale). The ~41K lines of deletion is a great cleanup. The TRX writer, filter builder, and AndroidTestAssemblyBuilder workaround are all clean.
Issues Found
| Severity | Category | File | Description |
|---|---|---|---|
| ❌ error | Nullable | TestInstrumentation.cs:338 |
Null-forgiving ! operator used — banned per repo rules |
| Patterns | Mono.Android.NET-Tests.csproj |
NativeAOT/LLVM $(ExcludeCategories) may no longer reach instrumentation via dotnet test |
|
| 💡 suggestion | Formatting | MyFragment.cs:17 |
#endif comment doesn't match #if __ANDROID_11__ |
Positive Callouts
- 👍 The
NumberOfTestWorkers=0comment is exemplary — explains why, not what, and documents the exact failure mode avoided. - 👍
AndroidTestAssemblyBuildercleanly works around NUnit'sAssembly.CodeBaseissue on Android single-file bundles. - 👍 CI yaml captures
adb logcat -d— this will save debugging time for native crashes. - 👍
NUnitCompatibility.csshim is pragmatic and well-documented with a clear removal plan.
Generated by Android PR Reviewer for issue #11224 · ● 42.4M
Comments that could not be inline-anchored
tests/TestRunner.Core/TestInstrumentation.cs:338
🤖 ❌ Nullable — ! (null-forgiving operator) is banned per repo conventions. Assembly.GetName().Name returns string?.
Replace with a null check:
=> inner.Build (assembly.GetName ().Name ?? throw new InvalidOperationException ("Assembly has no name"), options);Rule: Never use ! (null-forgiving operator)
tests/Mono.Android-Tests/Mono.Android-Tests/Xamarin.Android.RuntimeTests/MyFragment.cs:17
🤖 💡 Formatting — The #endif comment says ANDROID_11 but the #if guard is __ANDROID_11__. The comment should match the preprocessor symbol exactly for readability.
#endif // __ANDROID_11__Rule: #else/#endif comments must match the original expression (Postmortem #37)
tests/Mono.Android-Tests/Mono.Android-Tests/Mono.Android.NET-Tests.csproj:34
🤖 $(ExcludeCategories) MSBuild property defined here (for NativeAOTIgnore, SSL, LLVMIgnore, InetAccess, NetworkInterfaces) was previously wired through TestApks.targets → InstrumentationArguments → am instrument -e exclude. With the migration to dotnet test via MTP, the RunTestApp target is no longer invoked, so these MSBuild-only exclusions may not reach the on-device instrumentation.
The C# TestInstrumentation.ExcludedCategories handles `CoreCLR…
- TestInstrumentation.cs: replace ! null-forgiving with explicit null check + InvalidOperationException, per repo nullable rules. - MyFragment.cs: fix #endif comment to match #if __ANDROID_11__. - Mono.Android.NET-Tests.csproj + RuntimeTests/TestInstrumentation.cs: the MSBuild \ property was previously flowed to the on-device runner via 'am instrument -e' from the custom RunTestApp target. With 'dotnet test' + MTP that target no longer runs, so PublishAot/EnableLLVM exclusions never reached the device. Flow them through [AssemblyMetadata] instead, read at runtime in ExcludedCategories. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
jonathanpeppers
added a commit
to dotnet/java-interop
that referenced
this pull request
Jun 2, 2026
…eSetUp]/[OneTimeTearDown] (#1437) The legacy `[TestFixtureSetUp]` and `[TestFixtureTearDown]` attributes were removed in NUnit 3 in favor of `[OneTimeSetUp]` and `[OneTimeTearDown]`. They have been kept around in NUnitLite for back-compat, which is why the on-device Android test path used them under `#if __ANDROID__`. NUnitLite also supports the modern `OneTime*` names, so the conditional is no longer needed. Replacing the conditional with the unconditional modern attribute unblocks dotnet/android#11224, which migrates Mono.Android device tests from NUnitLite to stock NUnit 3.13.3 + MTP via `dotnet test`. Stock NUnit 3.13.3's `netstandard2.0` assembly does not contain the legacy attribute names, so consuming this code there currently requires a small compatibility shim (NUnitCompatibility.cs). After this change that shim can be deleted. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
dotnet/java-interop#1437 replaces the legacy [TestFixtureSetUp] / [TestFixtureTearDown] attributes with the modern [OneTimeSetUp] / [OneTimeTearDown] names that stock NUnit 3 ships out of the box, so the local NUnitCompatibility.cs shim is no longer needed. Submodule bump: e38211334 -> b982a1a3 (dotnet/java-interop): b982a1a [tests] Replace [TestFixtureSetUp]/[TestFixtureTearDown] with [OneTimeSetUp]/[OneTimeTearDown] (#1437) e68fa0a Bump to dotnet/android-tools@3c95718 (#1436) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Member
Author
|
/azp run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
Under stock NUnit 3 (after the dotnet test migration), an unhandled NUnit AssertionException thrown from inside a raw 'new Thread (...)' body becomes an unhandled exception on the .NET runtime side, which SIGABRTs the whole process. The instrumentation host then never gets a Finish() call and 'dotnet test' reports 'Zero tests ran' for the entire run -- masking the real failure (Assert #2 inside the worker) and taking out every other test in the assembly with it. Capture exceptions from the worker thread in a local, then rethrow on the test thread after Join() so an Assert failure (or any other exception) becomes a normal NUnit test failure instead of a process abort. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Member
Author
|
/azp run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
Build 14271469's NoAot lane reported 'Zero tests ran' with no diagnostics in logcat or the dotnet test output: process simply exited with code 1 mid-run. With no log of what threw, every flake turns into another CI iteration just to gather information. Hook AppDomain.UnhandledException and TaskScheduler.UnobservedTask- Exception so the next time something escapes from an NUnit runner / worker thread we at least learn which exception killed the process. The process will still terminate, but logcat will name the culprit. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Build 14271469 / APKs 1 / Mono.Android.NET_Tests-Release reproduced the exact NRE first diagnosed in 09acb6d: E DOTNET: Unhandled exception. System.NullReferenceException at NUnit.Framework.Internal.MethodWrapper.get_Name() at NUnit.Framework.Internal.TestExecutionContext.AdhocContext..ctor() at NUnit.Framework.Internal.Execution.TextCapture.WriteLine(String) at System.Console.WriteLine(...) at Java.InteropTests.JnienvTest.<>c.<ThreadReuse>b__7_0(IntPtr, IntPtr) F libc: Fatal signal 6 (SIGABRT) in tid 5229 (Thread-14) The 'minimize the diff' commit 4d924c4 reverted both the Console-to-Log conversion and the try/catch -- but only the try/catch needed to go. NUnit 3's Console.Out redirection is unsafe on a raw native pthread that has no TestExecutionContext, so route the diagnostics through Android.Util.Log to bypass the redirect entirely. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Member
Author
|
/azp run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
Build 14275720 had this test fail in 6 lanes simultaneously with two failure modes: #2 Expected: 36/37 But was: 37/38 (+1 extra peer at allocation) #3 Expected: 36/37 But was: 34/35 (1-3 fewer peers after GC) Build 14212378 on main (pre-PR) hit the same #2 failure, so this is a pre-existing flake -- not a regression. Before ac81297, mode #2 escaped from new Thread() and SIGABRT'd the process (manifesting as 'Zero tests ran' for the whole lane); the marshal-exception fix now surfaces it as a real test failure, which is why we suddenly see it. Runtime.GetSurfacedObjects() is process-global: NUnit/MTP per-test infrastructure (TestExecutionContext, listeners, Console capture) allocates and releases Java.Lang.Object peers around every test, so the count drifts by a few entries between snapshots. Observed drift across recent builds is in the range -3..+1; allow +/-5 tolerance which is plenty tight to still detect a real leak (which would dwarf this) while removing the noise. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…od-dotnet-test # Conflicts: # Directory.Build.props
Member
Author
|
/azp run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
Build 14283917 / Mono.Android.NET_Tests-IsAssignableFrom ran 262
tests and reported 3 JavaCast failures. That lane is configured in
build-tools/automation/yaml-templates/stage-package-tests.yaml as:
-p:IncludeCategories=Intune -p:_AndroidIsAssignableFromCheck=false
so it is meant to run ONLY tests tagged [Category("Intune")] (there
is exactly one such test) -- the 3 JavaCast tests legitimately
require the IsAssignableFromCheck feature switch to be on and would
always fail when the lane disables it.
Before this PR, `IncludeCategories` was carried via build-tools/
scripts/TestApks.targets into `am instrument -e include`, which the
old NUnitLite runner read at runtime. The migration to `dotnet test`
+ MTP removed that custom RunTestApp target, so the MSBuild property
was being silently dropped. Logcat confirmed: no `Including category:`
log lines fired for IsAssignableFrom (vs. multiple `Excluding category:`
lines for lanes whose subclass populates ExcludedCategories).
Flow IncludeCategories through the same [AssemblyMetadata] pipe we
already use for PublishAot / EnableLLVM:
* Mono.Android.NET-Tests.csproj emits
[AssemblyMetadata("IncludeCategories", "$(IncludeCategories)")]
when the property is non-empty.
* TestRunner.Core.TestInstrumentation gains a virtual
IncludedCategories property; BuildNUnitFilter merges it with any
legacy `am instrument -e include` extras into the existing
<or><cat>...</cat>...</or> include element.
* Xamarin.Android.RuntimeTests.TestInstrumentation overrides
IncludedCategories to read the attribute (refactored
HasAssemblyMetadata into a small GetAssemblyMetadata helper to
avoid two near-duplicate reflection loops).
Note: NativeAOT lane shows a related but distinct symptom -- the
Export filter is correctly emitted but Export_Method_* tests still
run, because NativeAOT trims [Category] attributes off the test
methods so NUnit can't match them. That is a separate, deeper issue
and not addressed here.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Migrate Mono.Android device test infrastructure from the legacy NUnitLite stack to
dotnet testwith the Microsoft Testing Platform (MTP), following the pattern established for theandroidtesttemplate in 031246d.Callers now run device tests with stock
dotnet test, no custom MSBuild targets required. NUnitLite is removed entirely from the workload — both as a framework assembly and as a vendored copy.Changes
New: MTP-based
TestInstrumentationbase classtests/TestRunner.Core/TestInstrumentation.csis now an MTP-based on-device runner using stock NUnit (NUnitTestAssemblyRunner) plusMicrosoft.Testing.Platform.androidtesttemplate: tests run synchronously on the instrumentation thread (NumberOfTestWorkers = 0, see comment in code), TRX is written to the app data dir, and the path is sent back through the instrumentation bundle so the host adapter can pick it up.Updated test projects
3.13.3from NuGet (no NUnitLite / TestRunner.NUnit), setIsTestingPlatformApplication=true, and use the newTestInstrumentationbase.Deleted (~41K lines)
src/Xamarin.Android.NUnitLite/— theXamarin.Android.NUnitLite.dllframework assembly (NUnitLite wrapper,TestSuiteInstrumentation, GUITestSuiteActivity/TestResultActivity/OptionsActivity, layouts).src-ThirdParty/NUnitLite/— the embedded NUnitLite source we used to compile into the above.tests/TestRunner.NUnit/—NUnitTestRunner/NUnitTestInstrumentation.tests/TestRunner.xUnit/—XUnitTestRunner/XUnitTestInstrumentation.Xamarin.Android.NUnitLite.dll/.pdb/.xmlremoved frombuild-tools/installers/create-installers.targets(no longer shipped in the framework).RunTestApp/ device-test MSBuild targets.The
XA4313deprecation message inCommon.targets(which suggested customers migrateXamarin.Android.NUnitLite→Xamarin.Legacy.NUnitLite) is preserved as the existing escape hatch; theXamarin.Legacy.NUnitLiteNuGet is published from a separate repo and is unaffected by these changes.CI updates (
apk-instrumentation.yaml)-t:Install, run withdotnet test --no-build --report-trx.adb logcat -dcapture so device-test crashes can be diagnosed without re-running CI.Test fixes / bookkeeping
NUnitCompatibility.csshim: NUnit 3.13.3'snetstandard2.0assembly does not include[TestFixtureSetUp]/[TestFixtureTearDown], but theJava.Interopsubmodule still uses them. The shim aliases them to[OneTimeSetUp]/[OneTimeTearDown]. Will be removed once [tests] Replace legacy [TestFixtureSetUp]/[TestFixtureTearDown] with [OneTimeSetUp]/[OneTimeTearDown] java-interop#1437 lands and the submodule is bumped.JavaSystem.LoadLibrary ("reuse-threads")inTestInstrumentation.OnCreateso it runs on the main Java thread instead of inside individual tests.ExcludedTestNames/ExcludedCategoriesto matchmain.Out of scope
EmbeddedDSOandLocale-Tests— old-style csproj that still<Reference Include="Xamarin.Android.NUnitLite" />. They need an SDK migration before this PR can fully delete that assembly's identity from the build (theXA4313deprecation will fire on these projects today). Tracked separately.Related
NUnitCompatibility.cs)UrlEscaping_Bug43411onmain(test had been silently dead under NUnitLite)