Rollup merge of #120322 - compiler-errors:higher-ranked-async-closures, r=oli-obk

Don't manually resolve async closures in `rustc_resolve`

There's a comment here that talks about doing this "[so] closure [args] are detected as upvars rather than normal closure arg usages", but we do upvar analysis on the HIR now:

cd6d8f2a04/compiler/rustc_passes/src/upvars.rs (L21-L29)

Removing this ad-hoc logic makes it so that `async |x: &str|` now introduces an implicit binder, like regular closures.

r? ```@oli-obk```
This commit is contained in:
Matthias Krüger 2024-01-26 06:36:39 +01:00 committed by GitHub
commit e400311486
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 29 additions and 29 deletions

View file

@ -4424,35 +4424,6 @@ impl<'a: 'ast, 'b, 'ast, 'tcx> LateResolutionVisitor<'a, 'b, 'ast, 'tcx> {
ExprKind::Type(ref _type_expr, ref _ty) => {
visit::walk_expr(self, expr);
}
// `async |x| ...` gets desugared to `|x| async {...}`, so we need to
// resolve the arguments within the proper scopes so that usages of them inside the
// closure are detected as upvars rather than normal closure arg usages.
//
// Similarly, `gen |x| ...` gets desugared to `|x| gen {...}`, so we handle that too.
ExprKind::Closure(box ast::Closure {
coroutine_kind: Some(_),
ref fn_decl,
ref body,
..
}) => {
self.with_rib(ValueNS, RibKind::Normal, |this| {
this.with_label_rib(RibKind::FnOrCoroutine, |this| {
// Resolve arguments:
this.resolve_params(&fn_decl.inputs);
// No need to resolve return type --
// the outer closure return type is `FnRetTy::Default`.
// Now resolve the inner closure
{
// No need to resolve arguments: the inner closure has none.
// Resolve the return type:
visit::walk_fn_ret_ty(this, &fn_decl.output);
// Resolve the body
this.visit_expr(body);
}
})
});
}
// For closures, RibKind::FnOrCoroutine is added in visit_fn
ExprKind::Closure(box ast::Closure {
binder: ClosureBinder::For { ref generic_params, span },