[TrimmableTypeMap] Add trimmable DotNetRun cases

[simonrozsival]

Adds trimmable typemap coverage to the DotNetRun device integration test, and fixes two issues in the trimmable typemap targets that were uncovered while bringing those scenarios up.

New test cases

Added to InstallAndRunTests.Get_DotNetRun_Data:

• Release CoreCLR + trimmable (UseMonoRuntime=false)
• Debug CoreCLR + trimmable (UseMonoRuntime=false)
• Release NativeAOT + trimmable

Fixes

dotnet run --no-build for trimmable CoreCLR

The trimmable _GenerateJavaStubs target listed _CleanIntermediateIfNeeded in its DependsOnTargets. On a full build that is redundant — the target already runs early via _AndroidBuildDependsOn in BuildOrder.targets. But during an incremental deploy that does not rebuild (the Install / DeployToDevice chain used by dotnet run without a build), the early Build chain never executes, so this was the only invocation of _CleanIntermediateIfNeeded — and it ran late, after _PrepareAssemblies had already populated obj/<config>/android/assets/<abi>/*.dll. It then deleted obj/<config>/android, and because _PrepareAssemblies had already executed it was skipped on the second pass and did not recreate the assemblies. _GeneratePackageManagerJava then failed with XAGCA7023 (DirectoryNotFoundException on assets/<abi>/Mono.Android.dll).

Removed _CleanIntermediateIfNeeded from the trimmable _GenerateJavaStubs DependsOnTargets so it matches the llvm-ir sibling target, which already omits it. Cleaning still happens early for full builds via BuildOrder.targets.

Parallel multi-RID race in post-trim Java source generation

CoreCLR trimmable builds with multiple RuntimeIdentifiers run per-RID inner builds in parallel. _GeneratePostTrimTrimmableTypeMapJavaSources previously ran inside each inner build and wrote to the same outer intermediate obj/<config>/typemap/linked-java directory, so inner builds raced cleaning and writing the same shared output. CI observed this as XAGTT7024 / "Directory not empty" while one inner build deleted the shared linked-java directory as another was using it.

The trimmed Java sources are RID-independent, so:

• Moved _GeneratePostTrimTrimmableTypeMapJavaSources to run once in the outer build, after _ResolveAssemblies has collected the per-RID ResolvedFileToPublish items, and before _GenerateJavaStubs / _CompileJava / _CompileToDalvik.
• Filter @(ResolvedFileToPublish) to a single representative RID (_PostTrimTypeMapFirstRuntimeIdentifier, the first entry in $(RuntimeIdentifiers)), so the target only consumes one RID's linked assemblies as input.
• Skip the target when running inside an inner per-RID build (_ComputeFilesToPublishForRuntimeIdentifiers != true).

Output remains in the outer intermediate directory where _GenerateJavaStubs expects it.

Fixes #11019
Related to #11604

[Copilot]

Pull request overview

Adds trimmable typemap coverage to the DotNetRun device integration test matrix, ensuring the trimmable typemap path is exercised under CoreCLR (UseMonoRuntime=false) in both Debug and Release configurations.

Changes:

• Add Release trimmable DotNetRun test case using AndroidRuntime.CoreCLR.
• Add Debug trimmable DotNetRun test case using AndroidRuntime.CoreCLR.

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[jonathanpeppers]

/review

[github-actions]

✅ Android PR Reviewer completed successfully!

[github-actions]

🤖 Code Review — ✅ LGTM (2 optional 💡 suggestions)

Solid, well-iterated change. Public CI is green, and the test-matrix additions are correct: trimmable is exercised for Release/Debug CoreCLR and Release NativeAOT, with Debug NativeAOT correctly omitted (NativeAOT is Release-only per IgnoreUnsupportedConfiguration).

Severity counts: ❌ 0 · ⚠️ 0 · 💡 2 (documentation only — non-blocking)

Worth flagging for reviewers (not a defect)

Despite the "add DotNetRun cases" framing, this PR also ships production build-logic changes that affect all trimmable CoreCLR builds, not just tests:

• _GeneratePostTrimTrimmableTypeMapJavaSources moves from AfterTargets="ILLink" → AfterTargets="_ResolveAssemblies", keys off @(ResolvedFileToPublish), and is gated on _ComputeFilesToPublishForRuntimeIdentifiers != 'true' so it runs once in the outer build (avoiding the per-RID race fixed in the de245ba/e400d73 commits).
• New _PostTrimTypeMapFirstRuntimeIdentifier + a first-RID filter to dedup multi-RID publishes.
• _CleanIntermediateIfNeeded dropped from the _GenerateJavaStubs override. I verified this is safe — it still runs early via $(_AndroidBuildDependsOn), and the override is now consistent with the Microsoft.Android.Sdk.TypeMap.LlvmIr.targets _GenerateJavaStubs, which also omits it.

These look correct and deliberate (they match the #11499 follow-up work), but the PR description doesn't mention them. A line in the description/changelog would help future archaeology.

Validation caveat

The new DotNetRun cases run on the device-integration pipeline, which isn't part of the green public CI checks visible here — so green public CI doesn't by itself prove the new cases pass on-device. The earlier resolved review thread confirms an on-device DotNetRun(True,"trimmable",CoreCLR) failure was the original motivation, so please confirm the on-device run is green before merge.

Nice work 👍

• Good incremental device coverage for the trimmable typemap.
• Correct handling of NativeAOT's Release-only constraint.
• Reuses the existing $([System.String]::Copy(...).Split(';')[0]) idiom already used for _TypeMapFirstAbi in the same file.

Two inline 💡 suggestions below — both are documentation-only and entirely optional.

Generated by Android PR Reviewer for issue #11486 · 1.1K AIC · ⌖ 29.7 AIC · ⊞ 37.6K
_{Comment /review to run again}

[jonathanpeppers]

I updated PR desc with copilot, merging.