[TrimmableTypeMap] Remove post-trim generation pass

[simonrozsival]

Summary

This PR partially backs out #11499, but intentionally keeps the part that fixed single-RID CoreCLR packaging.

#11499 mixed two separate changes:

Area from #11499 / follow-up cleanup	What this PR does
CoreCLR post-trim _GeneratePostTrimTrimmableTypeMapJavaSources target	Removes it
Second GenerateTrimmableTypeMap invocation after ILLink	Removes it
Task/generator switches used only by that second pass (JavaSourceInputDirectory, GenerateTypeMapAssemblies=false, CleanJavaSourceOutputDirectory, generateTypeMapAssemblies)	Removes them
typemap/linked-java Java source copy/filtering path	Removes it
XA4254/XA4255 docs/resources for the deleted Java copy path	Removes them
R8/acw-map post-trim assertions from the regression test	Removes/defer them to the separate R8 follow-up
Single-RID linked typemap assemblies added to ResolvedFileToPublish	Keeps it
ReadyToRun typemap assemblies replacing linked typemap assemblies for publish	Keeps it
Regression coverage that the APK packages linked/R2R typemap assemblies	Keeps it
Stale typemap/java/**/*.java outputs after the post-trim pass removal	Reconciles stale generated Java sources when the whole-world typemap generation target reruns, deletes matching stale android/src copies, and includes typemap/ in intermediate clean

The conceptual fix is: the trimmable typemap should be generated once before trimming and then reused. Running GenerateTrimmableTypeMap again after trimming invalidates the cache and defeats the point of the trimmable typemap design.

The Java-source cleanup is intentionally scoped to the current whole-world generator model: when _GenerateTrimmableTypeMap runs, the task has the full current desired Java source set, so files under typemap/java that were not emitted by that pass are stale and can be deleted. No-op incremental builds where _GenerateTrimmableTypeMap is skipped do not clean that directory.

The future per-assembly incremental split is tracked separately by #11494; that design will need a more explicit Java-output ownership/reconciliation model before this can become partial per assembly.

Are single-RID builds still fixed?

Yes. The single-RID fix from #11499 is still present: _AddTrimmableTypeMapToResolvedFileToPublish adds linked typemap assemblies to the publish pipeline, and _UseReadyToRunTrimmableTypeMapAssembliesForPublish swaps in R2R typemap assemblies when applicable.

The retained regression test builds a trimmable CoreCLR single-RID APK (RuntimeIdentifier=android-arm64, TrimMode=full, AndroidLinkTool=r8) and verifies the APK packages the linked/R2R typemap assemblies instead of stale pre-link typemap assemblies.

Validation

Previous validation:

• make prepare
• ./dotnet-local.sh build src/Xamarin.Android.Build.Tasks/Tests/Xamarin.Android.Build.Tests/Xamarin.Android.Build.Tests.csproj -c Debug -v:minimal
• ./dotnet-local.sh test bin/TestDebug/net10.0/Xamarin.Android.Build.Tests.dll --filter "Name=ReleaseCoreClrTrimmableTypeMap_SingleRuntimeIdentifier_PackagesLinkedOrReadyToRunTypeMapAssemblies"

Latest cleanup update:

• git diff --check
• Added regression coverage:
  • Execute_DeletesStaleGeneratedJavaSources
  • Build_WithTrimmableTypeMap_DeletesStaleGeneratedJavaSourcesAndCopies

Note: local make all still fails in unrelated Java.Interop CMake setup because clang++ cannot link libc++ (ld: library 'c++' not found).

[simonrozsival]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[Copilot]

Pull request overview

This PR removes the “post-trim” TrimmableTypeMap Java generation/copy pass (and its associated task switches + error codes/docs), returning to a single pre-trim typemap generation model while keeping the previously-added single-RID CoreCLR packaging fix and its regression coverage.

Changes:

• Removed the CoreCLR post-trim _GeneratePostTrimTrimmableTypeMapJavaSources target and the second typemap generation pass after ILLink.
• Simplified GenerateTrimmableTypeMap + TrimmableTypeMapGenerator by removing the “copy from input dir”/“skip assembly generation” modes and associated XA4254/XA4255 resources + docs.
• Updated the regression test to focus on verifying the APK packages the linked/R2R typemap assemblies (dropping the post-trim R8/acw-map assertions).

Show a summary per file

File	Description
src/Xamarin.Android.Build.Tasks/Tests/Xamarin.Android.Build.Tests/TrimmableTypeMapBuildTests.cs	Removes post-trim R8/acw-map assertions from the single-RID trimmable CoreCLR packaging regression test.
src/Xamarin.Android.Build.Tasks/Tasks/GenerateTrimmableTypeMap.cs	Removes post-trim copy/clean switches and always writes typemap assemblies + Java sources in a single mode.
src/Xamarin.Android.Build.Tasks/Properties/Resources.resx	Removes XA4254/XA4255 resource strings tied to the deleted post-trim Java copy path.
src/Xamarin.Android.Build.Tasks/Properties/Resources.Designer.cs	Removes the generated accessors for XA4254/XA4255.
src/Xamarin.Android.Build.Tasks/Microsoft.Android.Sdk/targets/Microsoft.Android.Sdk.TypeMap.Trimmable.targets	Removes linked-java path plumbing; always uses the pre-trim typemap Java output directory for stub copying.
src/Xamarin.Android.Build.Tasks/Microsoft.Android.Sdk/targets/Microsoft.Android.Sdk.TypeMap.Trimmable.CoreCLR.targets	Deletes the post-trim typemap Java generation target that ran after ILLink.
src/Microsoft.Android.Sdk.TrimmableTypeMap/TrimmableTypeMapGenerator.cs	Removes the optional “skip typemap assemblies” mode; generator always produces typemap assemblies.
Documentation/docs-mobile/messages/xa4254.md	Deletes docs for XA4254, now that the code path is removed.
Documentation/docs-mobile/messages/xa4255.md	Deletes docs for XA4255, now that the code path is removed.
Documentation/docs-mobile/messages/index.md	Removes XA4254/XA4255 from the error code index.

Copilot's findings

Files not reviewed (1)

• src/Xamarin.Android.Build.Tasks/Properties/Resources.Designer.cs: Language not supported

• Files reviewed: 9/10 changed files
• Comments generated: 1