Why, with a generic constraint on type parameter T of class P of “must inherit from A”, does the first call succeed but the second call fail with the type conversion error detailed in the comment:
abstract class A { }
static class S
{
public static void DoFirst(A argument) { }
public static void DoSecond(ICollection<A> argument) { }
}
static class P<T>
where T : A, new()
{
static void Do()
{
S.DoFirst(new T()); // this call is OK
S.DoSecond(new List<T>()); // this call won't compile with:
/* cannot convert from 'System.Collections.Generic.List<T>'
to 'System.Collections.Generic.ICollection<A>' */
}
}
Shouldn’t the generic constraint ensure that List<T> is indeed ICollection<A>?
This is an example of C#’s lack of covariance on generic types (C# does support array covariance). C# 4 will add this feature on interface types and also will update several BCL interface types to support it as well.
Please see C# 4.0: Covariance and Contravariance: