Making Functions working with Lazy<T> trimmer friendly

[bash]

I'm trying to make these two functions trimmer friendly:

public static class Lazy
{
	public static Lazy<TItem> Return<TItem>(TItem value) => new(value);

	public static Lazy<TResult> Select<T, TResult>(this Lazy<T> lazy, Func<T, TResult> selector)
		=> new(() => selector(lazy.Value));
}

dotnet publish with trimming enabled gives these warnings:

Lazy.cs(x,y): Trim analysis error IL2091: Lazy.Return<TItem>(TItem): 'T' generic argument does not satisfy 'DynamicallyAccessedMemberTypes.PublicParameterlessConstructor' in 'System.Lazy<T>'. The generic parameter 'TItem' of 'Lazy.Return<TItem>(TItem)' does not have matching annotations. The source value must declare at least the same requirements as those declared on the target location it is assigned to.
Lazy.cs(x,y): Trim analysis error IL2091: Lazy.Select<T,TResult>(Lazy<T>,Func<T,TResult>): 'T' generic argument does not satisfy 'DynamicallyAccessedMemberTypes.PublicParameterlessConstructor' in 'System.Lazy<T>'. The generic parameter 'TResult' of 'Lazy.Select<T,TResult>(Lazy<T>,Func<T,TResult>)' does not have matching annotations. The source value must declare at least the same requirements as those declared on the target location it is assigned to.

Is it OK to use [UnconditionalSuppressMessage] to suppress them?

I think it should be OK because:

1. I'm always creating Lazy<T> with either a value or valueFactory.
2. My code doesn't use the public parameterless constructor itself.

[vitek-karas]

I think in this case it's probably safe, but I would still suggest considering annotating the generic arguments of your methods to avoid the suppression. Suppressions are very tricky to get right and even if they are right at the time you add them, they may not be forever. If somebody changes the code to actually do call the public parameterless .ctor there will still be no warning (because of the suppression) and suddenly things will break.

I vaguely remember us discussing that some .ctors of Lazy don't need the annotation (as you mention), but currently there's no good way to express this fact in the language/attributes.

[bash]

Suppressions are very tricky to get right and even if they are right at the time you add them, they may not be forever.

Ah, that's a good point. Thank you!