Add test for EventHandlerType Invoke method reflection

[sbomer]

Adds testcase for #117200

@MichalStrehovsky I am not sure why this testcase doesn't reproduce the issue for ILC. But from debugging I see that:

• the delegate type gets an EETypeNode, not a ConstructedEETypeNode
• CreationAllowed returns true
• thus we never call MetadataManager.GetDependenciesDueToEETypePresence

It looks like #117194 will fix this, but I want to make sure I understand the intention. What's supposed to be the difference between CreationAllowed and ConstructedEETypeNode? Should we call GetDependenciesDueToEETypePresence for nodes where CreationAllowed is true but there's no ConstructedEETypeNode?

[Copilot]

Pull Request Overview

Adds a new test for invoking EventHandlerType.Invoke via reflection and updates attribute naming from Tool to By.

• Renames attribute parameters and property bindings from Tool to By across tests and utilities
• Adds EventHanderTypeUsedViaReflection test case and removes the old EventHanderTypeGetInvokeMethod file
• Updates Roslyn analyzer test to invoke the new test name

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
src/tools/illink/test/Mono.Linker.Tests/TestCasesRunner/ResultChecker.cs	Changed GetPropertyValue("Tool") to GetPropertyValue("By")
src/tools/illink/test/Mono.Linker.Tests.Cases/Reflection/EventHanderTypeUsedViaReflection.cs	Added new reflection test for EventHandlerType.Invoke
src/tools/illink/test/Mono.Linker.Tests.Cases/Reflection/EventHanderTypeGetInvokeMethod.cs	Removed obsolete test file
src/tools/illink/test/Mono.Linker.Tests.Cases/LinkXml/EmbeddedLinkXmlFromCopyAssemblyIsProcessed.cs	Switched Tool = to By = in [KeptTypeInAssembly] attributes
src/tools/illink/test/Mono.Linker.Tests.Cases/LinkXml/CanPreserveExportedTypesUsingRegex.cs	Switched Tool = to By = in link XML attribute
src/tools/illink/test/Mono.Linker.Tests.Cases/LinkXml/CanPreserveAnExportedType.cs	Switched Tool = to By = in link XML attribute
src/tools/illink/test/Mono.Linker.Tests.Cases.Expectations/Assertions/BaseInAssemblyAttribute.cs	Renamed Tool property to By
src/tools/illink/test/ILLink.RoslynAnalyzer.Tests/ReflectionTests.cs	Renamed test method to match new test class

Comments suppressed due to low confidence (2)

src/tools/illink/test/Mono.Linker.Tests.Cases/Reflection/EventHanderTypeUsedViaReflection.cs:21

• Typo in class name: "EventHanderTypeUsedViaReflection" should be renamed to "EventHandlerTypeUsedViaReflection" for consistency and clarity.

    public class EventHanderTypeUsedViaReflection

src/tools/illink/test/ILLink.RoslynAnalyzer.Tests/ReflectionTests.cs:38

• Typo in test method name: "EventHanderTypeUsedViaReflection" should be "EventHandlerTypeUsedViaReflection" to match corrected class name and filename.

        public Task EventHanderTypeUsedViaReflection()

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/illink
See info in area-owners.md if you want to be subscribed.

[jtschuster]

I am not sure why this testcase doesn't reproduce the issue for ILC.

Reproducing means the KeptMemberInAssembly("System.EventHandler.Invoke") assertion should fail, right? The assertions for KeptMemberInAssembly are in ResultChecker.cs, which looks like it doesn't run in NativeAOT tests.

[MichalStrehovsky]

I am not sure why this testcase doesn't reproduce the issue for ILC.

Reproducing means the KeptMemberInAssembly("System.EventHandler.Invoke") assertion should fail, right? The assertions for KeptMemberInAssembly are in ResultChecker.cs, which looks like it doesn't run in NativeAOT tests.

Yeah, I think this is running into the "What do you mean when you say the method/type is kept?"

In IL terms, it's easy. When we keep the IL metadata we say it was kept.

In native terms, it's not so easy. We can have the method body instruction, but not the metadata (name of the method, parameter list). We can have metadata for the method, but no method body. We can have a reflection-invisible MethodTable that is useful for casts, but cannot be looked up by name with Type.GetType. We can have metadata for a type, but no MethodTable. We can have metadata for a type, but only a limited MethodTable usable for casts.

The AOT bug here is that just having a delegate type in method signature doesn't mean we keep it to the extent of giving it a usable MethodTable. The Invoke method preservation code is waiting to see a usable MethodTable. We should have a usable MethodTable because someone could feed the delegate type into trim-safe reflection APIs (and also we have tons of suppressions in framework assuming Invoke method can be always found). Then we'll also have the Invoke method.