Anthony Latsis

@Catfish_Man @igormaka @slava_pestov @davedelong @harlanhaskins Does the string get resolved to an address using the mangler?

You know how 99.999999% of the time it’s not a compiler bug? Tonight it was a compiler bug. bugs.swift.org/browse/SR-16076 Major kudos to @harlanhaskins for helping me get more info from the compiler as it crashed.

@igormaka @slava_pestov @davedelong @harlanhaskins Oh so the ObjC runtime is ignorant of the type that a method ref was assigned in Swift, all it cares about is the symbol the ref resolved to. That makes a lot of sense.

@anthonylatsis @slava_pestov @davedelong @harlanhaskins As for the earlier example from the Rx libraries - there's a whole delegate proxy thing going on there that overrides `forwardInvocation:` and a whole lot of other magic.

@igormaka @slava_pestov @davedelong @harlanhaskins Turns out we get the expected unrecognized selector runtime exception in both cases, which makes me think the Objective-C runtime was either playing along or unaffected by this bug

@igormaka @slava_pestov @davedelong @harlanhaskins This caused a bit of a ruckus in regards to selector disambiguation, so I got curious and tested a selector against an instance with a missing implementation with both a 5.5 and 5.7 compiler

@slava_pestov @igormaka @davedelong @harlanhaskins Oh right there was a test with a selector of a force unwrapped ref to a dynamically looked up method that I had to add an expected error to in that PR — #selector(AnyObject.foo!)

@igormaka @slava_pestov @davedelong @harlanhaskins I expect them to have *always* crashed at compile time, but if you know of a case that used to somehow slip through, then yes, you wouldn’t have wanted to rely on its behavior.

@anthonylatsis @slava_pestov @davedelong @harlanhaskins In other words, the old way would produce an Optional that was never nil but would crash if applied to a class that doesn't implement the method?

@igormaka @slava_pestov @davedelong @harlanhaskins It makes sense to check for the existence of an implementation only once you have a conforming type to check against, so it wasn’t really a change of mind. An unbound ref to an optional req would crash prior to this change.

@anthonylatsis @slava_pestov @davedelong @harlanhaskins Unrelated to OP but could you please explain the rationale for changing from `((any P) -> (Int) -> Void)?` to `(any P) -> ((Int) -> Void)?` I just noticed this in the release notes (after a new compile error) and I'm curious.

@slava_pestov The brevity of the syntax is making me slightly nervous. What’s the secondary associated type going to be? Do you envision the angle brackets here as something that could be used for both type arguments and generic requirements?

@slava_pestov Yeah, some of them are obviously not. Why not Canberra though?

@slava_pestov Pretty cool, assuming all these are not vacation sites. Just realized I can name only 2 cities I’ve spent more than a couple months in

@ristkof @slava_pestov @TheOtherDaveS I second that. Though I’ve noticed they tend to be slightly less saturated when there’s no rooster involved

@harlanhaskins @chriseidhof @steipete I just realized the “func flatten<Wrapped>” example is not really a merit of SE-267. The proposal is about allowing “where” clauses on *non-generic* member declarations (generic ones already supported them at the time)

Anyone up for writing a Swift Evolution proposal? This seems useful and simple, kinda like the toggle() one from @chriseidhof

@jegnux @harlanhaskins Yep. Although this particular example will likely crash the compiler today. If I recall correctly, there was an older unresolved issue specific to same-type constraints in override signatures. If you were to replace T == Int with T: Comparable, dispatch should work as expected.

@anthonylatsis @harlanhaskins Thanks for your answer. It sounds evident now but it’s really weird to « override » a function without being able to call super. What about let x: Base<Int> = Derived<Int>() x.foo() Is it dynamically dispatched to Derived’s?

@jegnux @harlanhaskins The generic constraint will not allow you to call super.foo(). That paragraph is really only there to show that you *can* override such a member.

@harlanhaskins I’m surprised by the override here. What happens when calling `super.foo()` if T isn’t an Int?

@chernikovalexey @HaabHimself @kevin_reutter @KrauseFx You won’t be able to edit the time via the numpad unless you tap the selector anyway. How is that an unintuitive way to *discover* the dial?

Sorry, but how did we go from this to this in the last 13 years?

@wylsacom Так он и не разводит никого, прямо говорит – вот как я зарабатываю: переходите по ссылке и прогреваете мне душу пятихаткой

Стыдоба такое рекламировать. И желаю кренделю присесть за мошенничество быстро и решительно.

Два дня и сученыш расчехлился.

@QueenPans @meduzaproject Разве что если человек не лечиться из-за лени

@anthonylatsis @meduzaproject и даже в этих случаях

Ученые разобрали главные мифы об ожирении: — Чтобы похудеть, достаточно «есть меньше и двигаться больше» — Операция по уменьшению желудка — это «путь для ленивых» — Стыд помогает людям похудеть Все это неправда. meduza.io/feature/2020/0…