Under what conditions can you implement Copy for a union?

[Centril]

From #32836 (comment):

Under what conditions can you implement Copy for a union? For example, what if some variants are of non-Copy type? All variants?

@RalfJung notes in #55149 (comment) that:

The RFC doesn't say anything about that, but I think indeed the answer should be "all variants".

cc #55149

[petrochenkov]

the answer should be "all variants".

+1
If Copy is implemented for a union with non-copy fields, it undermines all the remaining move/initialization checking for unions.

[RalfJung]

I still think the same way.

unsafe impl Copy for semi-regularly comes up, but I do not see any reason to tie it to unions.

[Centril]

The current behavior does fortunately seem to enforce the "all variants" behavior:

#![feature(untagged_unions)]

struct NoCopy;

#[derive(Clone, Copy)]
struct IsCopy;

#[derive(Clone, Copy)]
union U {
    a: NoCopy,
    b: IsCopy,
}

results in an error:

error[E0204]: the trait `Copy` may not be implemented for this type
  --> src/lib.rs:8:17
   |
8  | #[derive(Clone, Copy)]
   |                 ^^^^
9  | union U {
10 |     a: NoCopy,
   |     --------- this field does not implement `Copy`

I agree that it is a sensible conservative default.

[varkor]

There don't seem to be any sensible alternatives to the current behaviour. What else could the condition possibly be?

[nikomatsakis]

@rfcbot fcp merge

We discussed this in today's @rust-lang/lang meeting and we all felt that this behavior seemed like the obvious and correct one. It is also the one that is presently implemented. Moreover, it's a conservative choice -- if we wanted to permit Copy impls under more conditions, we could do so at some later point. Therefore, I move that we "merge". (Since the behavior is already implemented, no changes are needed -- though one good question is whether we have suitable tests.)

[rfcbot]

Team member @nikomatsakis has proposed to merge this. The next step is review by the rest of the tagged team members:

• @Centril
• @cramertj
• @eddyb
• @joshtriplett
• @nikomatsakis
• @pnkfelix
• @scottmcm
• @withoutboats

No concerns currently listed.

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

See this document for info about what commands tagged team members can give me.

[petrochenkov]

The current behavior is not accidental and it has a test (union-copy.rs).