Miss Dada  🏳️‍⚧️

Probably the biggest way that self is getting special cased here is that it won't try to deref self but it will try to deref other arguments (why is `PartialEq<Inner> for Outer` fine there but not `PartialEq<Outer> for Inner`?)

but your example is already asymmetric, it's breaking because `y` is MyRc, not because it's not self. It fails if you swap y with x, it has nothing to do with the receiver being special cased

What's happening is that for x it's managing to find a single impl that doesn't require coercion. it's done! it's able to coerce for y because there's a single impl, so all the expected types are known

the moment you add the second impl, it no longer is able to find a single impl for x. which is usually fine, this means that &y needs inference to help pick, but deref coercions don't work with inference

Though i guess I see what you're saying: the way trait search works is via self, and yes, in that sense self is special cased, but that's not really special casing as much as it's fundamentally how traits work in the type system

> it's fundamentally how traits [have arbitrarily been chosen (which is fine) to] work FTFY ;)

right, i'm trying to say this isn't specific to deref coercions. (and also that it's not the same rules as .bar() at all, autoderef works differently from deref coercions)https://twitter.com/sgrif/status/1199080936320557056 …

The rules I assume we'd use are basically an extension of that, but I def can't fit anything more in depth than "fewest derefs left to right wins" without writing a formal RFC (which I'd love to do but probably won't any time soon)