Re: [PATCH v4 21/27] rust: types: add `ForLt` trait for higher-ranked lifetime support

[Eliot Courtney]

On Sun May 24, 2026 at 12:46 AM JST, Danilo Krummrich wrote:
> On Fri May 22, 2026 at 1:34 AM CEST, Danilo Krummrich wrote:
>> [ Handle macro_rules! invocations in the ForLt! proc macro's covariance
>>   and WF checks.
>>
>>   Proc macros cannot expand macro_rules! invocations, so the
>>   syn::Visit-based has_lifetime() and replace_lifetime() helpers cannot
>>   inspect types hidden behind macro calls. This caused the covariance
>>   proof to be silently skipped and lifetime substitution to fail for
>>   such types.
>>
>>   Add an explicit Type::Macro arm to Prover::prove to conservatively
>>   require a compiler-assisted covariance proof. Detect macro-containing
>>   types with has_macro() and use a WithLt trait projection for lifetime
>>   substitution instead of AST-level replacement. - Danilo ]
>
> I was made aware by Gary that WFness is not actually being checked, which is why
> the static expansion was needed in the first place.
>
> Thus macros will remain unsupported (even covariant macros are rejected), yet we
> can improve the error message by adding the following diff on top of the
> previous v3 patch.
>
> diff --git a/rust/macros/for_lt.rs b/rust/macros/for_lt.rs
> index df2027789713..75d0ce450bf0 100644
> --- a/rust/macros/for_lt.rs
> +++ b/rust/macros/for_lt.rs
> @@ -107,6 +107,12 @@ fn visit_lifetime(&mut self, lifetime: &Lifetime) {
>                      self.1 = true;
>                  }
>              }
> +
> +            // Macro invocations are opaque; conservatively assume they may
> +            // reference the lifetime.
> +            fn visit_macro(&mut self, _: &syn::Macro) {
> +                self.1 = true;
> +            }
>          }
>
>          let mut visitor = HasLifetime(lt, false);

With this change,

Reviewed-by: Eliot Courtney <ecourtney@xxxxxxxxxx>

I think also this implementation considers &'a mut dyn Trait to be
covariant in 'a which isn't the case in general IIUC, but I can't see
any way it could cause an issue currently.