Check if generic types encode its arity when using GetDisplayName

[mateoatr]

We currently assume that generic type names always encode its arity in the IL, whereas the ECMA explicitly states that both the grave accent and arity are optional.

Fixes #1435 and #1448

[vitek-karas]

Did you find a repro for this, or is it more of a theoretical thing?
I'm not questioning the fix, I'm trying to figure out if this is something we should try to take for .NET 5, or if it can wait for later release.

[mateoatr]

Did you find a repro for this, or is it more of a theoretical thing?

I opened this because of #1435, although I couldn't repro it in my machine (hitting other issues before getting to this one).

[MichalStrehovsky]

I opened this because of #1435, although I couldn't repro it in my machine (hitting other issues before getting to this one).

Could the undecorated thing be reachable for non-generic nested types nested under a generic type?

class Gen<T>
{
    // This one is "A" with one generic parameter ("inherited" from the containing type)
    class A { }

    // This one is "B`1", with two generic parameters (one is "inherited" from the containing type)
    class B<U> { }
}

[mateoatr]

Yes. We have tests for both scenarios already:

https://github.com/mono/linker/blob/7eae638feafc7dc766f51f440e760220707a87ad/test/Mono.Linker.Tests/Tests/GetDisplayNameTests.cs#L156-L157

https://github.com/mono/linker/blob/7eae638feafc7dc766f51f440e760220707a87ad/test/Mono.Linker.Tests/Tests/GetDisplayNameTests.cs#L122-L123

[mateoatr]

I'm trying to figure out if this is something we should try to take for .NET 5, or if it can wait for later release.

Based on #1435 and #1448, I think we should take it for .NET 5.

[vitek-karas]

OK - can you please prepare a PR to port this to the 5 release branch as well?