@badcel @khalidabuhakmeh I guess I'm not fully understanding what #nullable does then, since it clearly does not mean things can't be null, at least not in all contexts.

I did use `where T: notnull` now and the error also goes away. But in my mind, that should be the default when #nullable is enabled.

@khalidabuhakmeh But yes, nonetheless, declaring it nonnull does fix the problem.

`class Something<T> where T : notnull`

But I still don't understand why. It seems to go against what `#nullable enable` means.

@khalidabuhakmeh But I have `#nullable enable` on the file, implying all references are non-null unless declared otherwise, like `T?`.

Also, if I mark source as `T? source` then I get errors in my code, implying that it isn't actually nullable, except in this circumstance.

@andrewlock I see you enabled nullable, which is quite a minefield with a source-only package that wants to be able to work not only with nullable on or off, but also in a language where nullable isn't an option.

You end up with stuff like this:

https://github.com/xunit/assert.xunit/blob/309647ac2fc29997937a22afaaeb0a73fb80d14b/CollectionAsserts.cs#L1-L8

https://github.com/xunit/assert.xunit/blob/309647ac2fc29997937a22afaaeb0a73fb80d14b/CollectionAsserts.cs#L457-L472

You can't just "#nullable disable" because the compiler might not support it. So if they enable nullable but forget to "tell you", then you have to turn off a bunch of nullable "violations".