Slices don't implement copy.

[mahkoh]

[T]: Copy where T: Copy

does not hold even though it could. Together with size_of_val, this would allow one to write various functions generic over all Copy types whether they are sized or unsized.

[mahkoh]

It's seems like Copy requires Sized but i don't see a reason for that either.

[eddyb]

See #25776.

[mahkoh]

The case where I noticed this is the following:

trait Pod: Copy { }

That is,

• Pod types can always safely be copied,
• Pod types don't have to be unsized.

Both statements are true (given the definition of Pod used in lrs, unrelated to the Pod mentioned in the linked issue.) But I had to remove the Copy bound because I couldn't implement

impl<T: Pod> Pod for [T]

which is necessary because I want to implement

impl<T: Pod+?Sized> AsMut<[d8]> for T

The comment by @nikomatsakis seems to mostly address the cases where you're using Copy+?Sized directly, that is, where you care about Copy. But in the AsRef implementation I only care about the Pod aspect.

It seems that the built-in Sized bound can be removed in a backwards compatible way by adding an implicit Sized bound in libcore (should already exist via the Clone bound). If you then allowed Copy to be (explicitly) implemented for slices, the code above would work.

[cyplo]

Any news on this one? Do we need RFC for it ?

[nikomatsakis]

I am not sure whether I think we should permit sized types to be copy. I outlined some of my thoughts in this comment, which @eddyb already linked to. However, in the meantime, I've started to wonder if we should support moving and copying of DSTs just by using allocas at codegen time, which would allow us to make them work much more uniformly, almost like any other type. That would also make it fine for unsized types to be copy, though it might lead to footguns in some cases, but no different than a type like [u32; 1024] being Copy).

[steveklabnik]

Triage: given niko's last comment, I think I'm going to close this one too. It's been a few years, and this feels very much like a "needs an RFC" thing to me. Let me know if that's wrong. Thanks!