Empty bounds lists in where clauses cause the type to not be WF-checked.

[eddyb]

This produces no error:

fn foo() where Vec<dyn Copy>: {}

Just like the other known "missing WF checks" bug (#21903):

type Foo = Vec<dyn Copy>;

Unlike type aliases, I don't believe you can even reach the type written in the where clause without any bounds, since it doesn't even "remain in the type/trait-system" (e.g. inference can't pick it up).

So it's not as bad, but I did come up with use for it as a WF requirement, in #44580 (comment).

cc @rust-lang/lang @RalfJung @petrochenkov

[Centril]

Why is fn foo() where Vec<dyn Copy>: {} even passing the parser?
There is no bound on the RHS of :...

[cramertj]

@Centril We're pretty permissive with parsing "zero or more" of many syntax elements, e.g. fn foo() where {} also compiles.

[Centril]

@cramertj Oh I see; that's where this is coming from. It didn't click initially; Thanks! ❤️

[Centril]

We discussed this briefly on the meeting and the consensus was that if crater passes, then this is fine.

[eddyb]

This is waiting on @nikomatsakis or someone else r+'ing #53707 (crater was clean).