I’ve found a strange inconsistency between the nullability of a variable in the hover-over tooltip in Visual Studio / sharplab.io and the compiler’s handling of the variable re nullability.

I wanted to check if I’m missing something or if this is maybe a bug…

Live link: sharplab.io

#nullable enable

using System;
using System.Collections.Generic;
using System.Linq;

public sealed class RectangleInfo
{
}

public sealed class PreviewLayout
{
    public sealed class Builder
    {
        public Builder()
        {
            this.Screens = new();
        }
        public List<RectangleInfo> Screens
        {
            get;
            set;
        }
        public PreviewLayout Build()
        {
            return new PreviewLayout();
        }
    }
}

public static class LayoutHelper
{
     public static PreviewLayout GetPreviewLayout()
     {
         var builder = new PreviewLayout.Builder(); 
         var screen = builder.Screens.Single();
         Console.WriteLine(screen.GetType().FullName);
         return builder.Build();
     }
}

If you hover the mouse over the screen in var screen = builder.Screens.Single(); you’ll see this tooltip:

(local variable) RectangleInfo? screen

(note that RectangleInfo? is nullable, which is strange because Single() definitely returns a RectangleInfo in this case and there’s no other nullability warnings / errors)

but in the next line it’s being treated as “not null” because this gives no nullability warnings:

screen.GetType().FullName

If you replace var in var screen = ... with RectangleInfo the tooltip reports:

(local variable) RectangleInfo? screen

I get the same behaviour in Visual Studio, so presumably this is something the C# compiler / Roslyn is doing.

Have I missed something that means screen should legitimately be a nullable RectangleInfo??