Self as default type isnt typechecked

[DutchGhost]

The following code compiles:

struct B<P: Sized = [Self]>(P);

but shouldn't, because [Self] is NOT Sized.

If we change [Self] with a [u8], like this:

struct B<P: Sized = [u8]>(P);

it fails to compile, because [u8] isnt sized.

Also note that if we would write impl B {} in the case of P being of default type [Self], we get an ICE:

Related to #59956 (comment)

[DutchGhost]

Using godbolt, this slipped in with 1.32, where Self is allowed in type defs: #56366

[Centril]

cc @alexreg

[alexreg]

Thanks for the report.

@Centril I think the straightforward solution is to ban mentions of Self as type parameter defaults. After all, if you replace Self with B in the above example, you get:

error[E0391]: cycle detected when processing `B::P`

Sound fair?

[Centril]

That's a breaking change; consider e.g.:

enum L<P = Self> {
    N,
    C(u8, Box<P>),
}

fn main() {
    L::C(0, Box::new(L::<()>::N));
}

I think the right thing to do is to account for Self when checking the bounds of a type definition.

[alexreg]

Oh, I didn't realise Self was previously permitted...

[Centril]

This appears to be an instance of a wider problem:

trait Foo<Bar: Sized = [Self]> {}

Has compiled, but shouldn't have (?), since 1.0.

cc @nikomatsakis

[alexreg]

Thoughts after discussing this with @Centril: maybe we should be allowing this after all (or have no other choice due to the way the type system works), and should instead put in checks on things like imp B { ... } (with no binding for P) and possibly elsewhere.

[pnkfelix]

triage: P-high. Leaving nominated for discussion at rustc meeting. Leaving unassigned for now.