Implementing `Copy` can be a breaking change

[Jules-Bertholet]

I tried this code:

#[derive(Clone)]
struct Foo<'a>(&'a ());

// Uncomment this line and the code stops compiling
//impl Copy for Foo<'static> {}

struct Bar<'a>(Foo<'a>);

fn do_thing<'a>(b: Bar<'a>) {
    drop(b.0);
}

I expected to see this happen: Uncommenting the line in question does not break the code, as implementing Copy should not be a breaking change

Instead, this happened:

error: lifetime may not live long enough
  --> src/lib.rs:10:10
   |
9  | fn do_thing<'a>(b: Bar<'a>) {
   |             -- lifetime `'a` defined here
10 |     drop(b.0);
   |          ^^^ copying this value requires that `'a` must outlive `'static`

Meta

Regressed in 1.42 according to Godbolt.

@rustbot label A-borrow-checker A-lifetimes A-traits T-types regression-from-stable-to-stable

[QuineDot]

Intentionally denied since Rust 1.41.1.

Here's the PR.

[Jules-Bertholet]

Intentionally denied since Rust 1.41.1.

The example from the blogpost has two drops, which of course must require the field type to be Copy, because otherwise you would access a field that has already been moved out of. My example has only one drop, which should not require any such restriction. So it seems the fix PR was overly strict.

[Noratrieb]

Adding a copy impl for a specific lifetime is not something that should ever be done, and a bug in the program.

[BoxyUwU]

Removing the regression labels since this was intentional (if you disagree feel free to re-add)