On Tue, Apr 1, 2025 at 5:41 AM Michael Matz <m...@suse.de> wrote: > > Hello, > > On Mon, 31 Mar 2025, Bill Wendling wrote: > > > 1. The use of '__self' isn't feasible, so we won't use it. > > That's a bold statement. How's that? The only thing I read here is that > the very spelling of "self" was objected to. So, call it _Self, > _Selfref, or something. Even _Whatever42 would be better. > > Lookup rule changes only for identifiers used in attributes (all of them? > some of them?) is ... well, not good language design? I think you used > the word "overreaching" upthread, correctly so. And that for wanting to > avoid the established member-lookup syntax A.B by way of "we won't use > it"? That seems a terrible tradeoff. I don't understand the push for > that approach, at all. > It's not the name but the use of the identifier itself. If we can't enforce using it in every situation it's no longer useful. This is something that Apple (and the Clang community in general) is very against doing (not the identifier, but requiring its use). I still agree that adding a new scoping rule is overreach, but I'm trying to compromise so that we don't have multiple compilers with competing syntaxes. And as Yeoul pointed out in her RFC, we already have a similar lookup scheme with the "guarded_by" attribute.
-bw
