trait_mod:<is>(Routine:D, |) is called on the proto, not the candidate

[gfldex]

For a multi trait_mod:<is>(Routine:D, |) is always called on the proto, never on the actual candidate. This has been deemed wrong in the discussion: https://irclogs.raku.org/raku-dev/2023-06-05.html

    multi sub trait_mod:<is>(Routine \r, :$renamed) { r.candidates.say; r does role :: { method name { $renamed } } }

    proto sub foo(|) is renamed<foo-proto> { * }
    multi sub foo(Int $) { say &?ROUTINE.name }
    multi sub foo(Str $) is renamed<stringer> { say &?ROUTINE.name }
    multi sub foo(Failure) is renamed<failer> { fail('oi‽') }

    foo(42);
    foo('answer');
    # foo(Failure);

Expected Output

foo
stringer

Actual Output

foo
failer

Environment

• Operating system: Linux dexhome 6.1.0-9-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.27-1 (2023-05-08) x86_64 GNU/Linux
• Compiler version (perl6 -v or raku -v): Welcome to Rakudo™ v2023.05-98-ge8b477b58. Implementing the Raku® Programming Language v6.d. Built on MoarVM version 2023.05.

[jnthn]

It's called on the candidate, but roles are not closures (and ideally we'll find a way to make this spit out a warning in the new compiler frontend), so $renamed in the method always references the value in the most recent call of the trait.

[gfldex]

Indeed, I stepped into the "not quite golfed enough"-trap.

    multi sub trait_mod:<is>(Routine \r, :$thiscalled) { dd r; }

    proto sub foo(|) is thiscalled<?> { }
    multi sub foo(Str $s, :$isa = <stringer>) is thiscalled<?> { }
    multi sub foo(Int $s, :$isa = <inter>) is thiscalled<?> { }

OUTPUT:

Sub foo = proto sub foo (|) { #`(Sub|5551856403328) ... }
Sub foo = multi sub foo (Str $s, :$isa = "stringer") { #`(Sub|5551856429696) ... }
Sub foo = multi sub foo (Int $s, :$isa = "inter") { #`(Sub|5551856430848) ... }

Do we have an issue-tag for "needs a warning"?