With this PR the output from Mono is now

typeof(delegate*<int>):				System.MonoFNPtrFakeClass
typeof(delegate*<int>[]:			System.MonoFNPtrFakeClass[]
typeof(delegate*<int>*):			System.MonoFNPtrFakeClass*
typeof(delegate*<int>[]).GetElementType():	System.MonoFNPtrFakeClass

Methods on Foo:
Void Method(System.MonoFNPtrFakeClass)
 System.MonoFNPtrFakeClass
Void Method(System.MonoFNPtrFakeClass)
 System.MonoFNPtrFakeClass

Compare with CoreCLR which outputs System.IntPtr and fnptr in various places (see https://twitter.com/MStrehovsky/status/1311247635756412928/photo/2)

I'm going to take take a look over the rest of our code to see if there's a MONO_TYPE_FNPTR case missing elsewhere.

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

Grepped through the Mono code for MONO_TYPE_PTR and looked to see if there was a MONO_TYPE_FNPTR case nearby.

There are a few places where we could do better, but I don't think we'd crash:

• in the debugger-agent, buffer_add_value_full and buffer_add_fixed_array and some other places don't handle function pointers and they end up in NOT_IMPLEMENTED which won't crash the runtime.

Some that I'm not sure about:

• interpreter build_args_from_sig doesn't have a FNPTR case and asserts.
• dwarfwriter get_type_die returns I4 for a fnptr (for regular pointers it returns I)
• dwarfwriter emit_type doesn't emit anything for fnptr (but also doesn't emit anythign for ptr)
• AOT compiler encode_type doesn't have a FNPTR case and asserts
• mini-runtime create_runtime_invoke_info doesn't have a FNPTR case and asserts
• object.c invoke_array_extract_argument doesn't have a FNPTR case and asserts
• SRE encode_type doesn't know how to encode a FNPTR type and will assert

A few that might cause problems:

• object.c mono_field_get_value_object_checked doesn't have a FNPTR case and asserts
• marshal-ilgen.c I think struct marshalling probably has missing cases, which will assert
• RuntimeFieldInfo.SetValueInternal doesn't have a case for FNPTR and will assert
• inflate_generic_type doesn't have a FNPTR case and so doesn't inflate delegate *<S,T> where S and T are generic parameters. This one seems bad, but creating some generic struct with a fnptr field and using it seems to work:

  public unsafe struct S<T> {
     public delegate *<T,T> fptr;
  }
  static int F (S<int> s, int x) => s.fptr (x);
  ...
  {
     S<int> s;
     s.fptr = &IntHelper.TimesTwo;
     Console.WriteLine (F (s, 42)); // prints 84, as expected
  }