Add regression test for #129541

[zachs18]

(maybe?) closes #129541 by adding a test that the code in question continues to compile.

[rustbot]

r? @fee1-dead

rustbot has assigned @fee1-dead.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[fee1-dead]

I'm putting my review on hold because I'm unsure about the implications of this. This is a concrete example of something that went legal after #122493 landed, when the PR itself did not anticipate allowing more code to compile ("This should not make more or less code compile"). The code snippet here looks fine and sound to me; but I'm not sure if there would be some cursed code that would be unsound and then allowed by the PR.

cc @lukas-code @rust-lang/types

[compiler-errors]

I'll take a look at this, or if I forget, then lcnr can. As I noted in #129541, I'd like to know precisely why #122493 caused this to compile, although I generally don't believe this to be a regression per se.

[compiler-errors]

r=me when the fcp is done in the linked issue

[lukas-code]

Please also add the test from #129541 (comment) (or a minimized version):

#[derive(Clone)]
struct Hello {
    a: <[Hello] as ToOwned>::Owned,
}

so that we actually test that this is allowed because Sized is coinductive and not just because we skip the Sized check for enums now.

[compiler-errors]

@bors r+ rollup

[bors]

📌 Commit 5f72f9d has been approved by compiler-errors

It is now in the queue for this repository.