Why does C#/CLR not support method override co/contra-variance?

Question

There are quite a few questions & answers about hacking around the limitation of C# not allowing method return (and argument) types to be changed to compatible types on overrides, but why does this limitation exist, either in the C# compiler or in the CLR? As I an see, there is nothing that could break if co/contra-variance was allowed, so what is the reasoning behind it?

A similar question could be asked for widening access parameters - eg overriding a protected internal method with a public method (something which Java supports, IIRC)

Answer 1

It does, you just have to wait for VS2010/.Net 4.0.

Answer 2

To expand on Joel's answer - the CLR has supported limited variance for a long time, but the C# compiler doesn't use them until 4.0 with the new "in" and "out" modifiers on generic interfaces and delegates. The reasons are complicated, and I would get into a mess trying to explain, but it isn't as simple as it seems.

Re making a "protected internal" method into a "public" method; you can do this with method hiding:

public new void Foo(...) { base.Foo(...); }

(as long as the arguments etc are all also public) - any use?

Answer 3

Eric Lippert already answered this way better than I could.

Check out his series on Covariance and Contravariance in C#

and

How does C# 4.0 Generic Covariance & Contra-variance Implmeneted?

EDIT: Eric pointed out that he doesn't talk about return type convariance but I decided to keep the link in this answer because it is a cool series of articles and someone might find it useful if looking up this topic.

This feature has been requested and almost 5 years ago Microsoft has responded with "Thanks for logging this. We hear this request a lot. We'll consider it for the next release."

And now I'll quote Jon Skeet because it would not be a proper answer on StackOverflow without an answer by Jon Skeet. Covariance and void return types

I strongly suspect that the answer lies in the implementation of the CLR rather than in any deep semantic reason - the CLR probably needs to know whether or not there's going to be a return value, in order to do appropriate things with the stack. Even so, it seems a bit of a pity, in terms of elegance. I can't say I've ever felt the need for this in real life, and it would be reasonably easy to fake (for up to four parameters) in .NET 3.5 just by writing a converter from Func to Action, Func to Action etc. It niggles a bit though :)