r? @nnethercote

rustbot has assigned @nnethercote.
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

Seems reasonable to me but I will defer to someone who actually knows about lowering, in case I'm overlooking something.

r? @fmease

the unstable fn_delegation feature duplicates ast nodes, causing the lowerer to visit the same generic parameters multiple times and triggering an ice on a duplicate def_id assertion.

Just curious, why the following won't make ICE if we will lower generic parameters multiple times for delegation items🤔:

#![feature(fn_delegation)]
#![allow(incomplete_features)]

fn main() {
    trait Foo {}
    fn foo<N: for<'a> Foo>() {}
    reuse foo;
}

The problem here is that in current implementation we copy const type and eventually lower it twice. I don't think that the removal of the duplicated copy assertion is a correct way to fix this problem, this assertion did it job and highlighted the real issue. This issue should be solved by a proper const type propagation from HIR to TY level (i.e., adding stubs in hir::Ty in InferDelegation variant).

Just curious, why the following won't make ICE if we will lower generic parameters multiple times for delegation items🤔:

In this case N is a generic type parameter and for<'a> Foo is a predicate, predicates are not created at AST -> HIR lowering stage, they are inherited later.

thanks @aerooneqq for the explanation! that makes total sense.

however, i feel like implementing proper const-type propagation across the hir/ty level is a bit too big in scope for this single pr.

would it make sense to leave this as a temporary stopgap to fix the ice for now, or is it better to just close this and track the broader architectural fix in a separate issue?

I think the second option is preferable, as there is at least one more ICE (#153499) that relates to const types. Now we are researching a better way to work with generics and delegation code generation at all, If this prototype will be merged then it will be much easier to fix those ICEs, so I would wait until we finish with this research PR.