[mono] Support function pointer reflection introspection API

[ivanpovazan]

This PR implements support for function pointers reflection introspection API with Mono based on the design doc.

Done

• Added support for IsFunctionPointer
• Added support for IsUnmanagedFunctionPointer
• Added support for ToString
• Added support for GetFunctionPointerParameterTypes
• Added support for GetFunctionPointerReturnType
• Added support for GetRequiredCustomModifiers
• Added support for GetOptionalCustomModifiers
• Added support for GetFunctionPointerCallingConventions
• Implement ModifiedType.Mono
  • Added support for GetModifiedParameterType
  • Added support for GetModifiedFieldType
  • Added support for GetModifiedPropertyType
• Enabled previously disabled function pointer tests
• Added additional tests to src/libraries/Common/tests/System/ModifiedTypeTests.cs to verify fetching custom modifiers attached to generic arguments exposed via ModifiedType on a generic field and property (this is an addition to existing generic parameter tests

Not done

What is missing in this PR is proper support for function pointer type equality and type compatibility which was reported as a separate issue: #90308 and will be handled separately.

The remaining tests cases verifying this missing functionality have been updated with the new tracking issue.

---

Fixes #71095

[ivanpovazan]

@vargaz, I have a question regarding fptr namespace and name handling.

The tests expect that:

runtime/src/libraries/Common/tests/System/FunctionPointerTests.cs

Lines 27 to 28 in 75ecb1a

	Assert.Equal(string.Empty, t.Name);
	Assert.Null(t.Namespace);

Should we keep internally the following values:

runtime/src/mono/mono/metadata/class-init.c

Lines 1598 to 1599 in fa9ad22

	result->name_space = "System";
	result->name = "MonoFNPtrFakeClass";

and only special case ves_icall_RuntimeType_GetName and ves_icall_RuntimeType_GetNamespace for function pointers?

I am asking this because changing result->name_space = NULL; would break in different places as we assume name_space is not NULL.

[vargaz]

Special casing GetName/GetNamespace would probably cause less problems than setting the namespace to NULL.

[ivanpovazan]

Not sure if this is the best place to root the assembly in question, but it needs to be rooted to prevent failures in EnableAggressiveTrimming setups for FPTR testing.

PS The assembly is used in several places to test FPTR support.

[lambdageek]

LGTM, although I'd like TypeSignature to contain more of the logic about figuring out which signature source is being used so that less of it leaked out into the rest of ModifiedType

[ivanpovazan]

/azp run runtime-extra-platforms

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[ivanpovazan]

The CI failures are unrelated.

[lambdageek]

LGTM thanks!

[lambdageek]

👍 This whole file looks much clearer now, thanks!

[steveharter]

Yes and glad the template worked for mono.

[steveharter]

Thanks for adding these tests.

[steveharter]

I assume mono_class_create_fnptr() is still used along with an internal type of "MonoFNPtrFakeClass"?

[steveharter]

Thanks - the Mono integration looks fairly clean.

[ivanpovazan]

I assume mono_class_create_fnptr() is still used along with an internal type of "MonoFNPtrFakeClass"?

Yes, that is correct.