Arc and Weak should be #[repr(transparent)]

[SoniEx2]

I make these assumptions: https://docs.rs/eventbus/0.2.0/src/eventbus/lib.rs.html#155

And I won't stop making them, even if I technically can't rely on them. There's absolutely no reason why you can't compare an Arc and a Weak without upgrading.

[NeoLegends]

(ignoring the highly offensive comment in your code that suggests you don't even want help)

Why is temporarily upgrading for the comparison a problem? Did you run into measurable performance issues?

That said, I don't think they should be #[repr(transparent)], because that would expose too much implementation detail. A proper way to see if an Arc and a Weak point to the same thing is the better option here.

[jonas-schievink]

Yeah, this could be much more easily solved by adding Weak::as_ptr, and possibly Arc/Rc::as_ptr (which you don't really need since those Deref already). Alternatively, adding all 3 missing combinations of Arc/Rc::ptr_eq would work too.

[SoniEx2]

If you have an &Arc and an &Weak, e.g. because the Arc is the owned variant and the Weak is in a larger datastructure like a HashMap or a Vec, you can avoid atomic operations if you can just compare the Arc and the Weak directly. I'm not sure when it's desired to avoid atomic operations but ppl tell me it's a good idea to avoid atomic operations whenever possible. (they also say that, when possible, Relaxed is better than Aquire/Release is better than SeqCst)

Newer versions of eventbus don't compare Arcs and Weaks. This may change in the future, tho.

[Mark-Simulacrum]

We provide as_raw/into_raw on Weak and Arc (Rc as well, I believe) through #60728.