I get that for macOS as well when executed in its own script file. Does seem a little weird. @jckarter, @slavapestov, any ideas?

Did it get autoreleased? It's unlikely, but the compiler has the theoretical right to push the release past the imminent end of the program as well.

Well, I get the same output with this:

import Foundation

var foo: AnyObject?
weak var fooWeak: AnyObject?
autoreleasepool {
  foo = "foo" as NSString
  fooWeak = foo
  print("before: \(foo) / \(fooWeak)")
  foo = nil
}
print("after: \(foo) / \(fooWeak)")

@gparker42, is there anything different about weak refs in the simulator that you're aware of?

You're probably being fooled by some immortal string optimization such as tagged pointers or string ROM. Try string values that are longer and less likely to be shared.

I get the following behavior from Joe's arrangement on macOS:

before: Optional(foo) / Optional(foo)
after: nil / Optional(foo)

before: Optional(foo1234!@#$) / Optional(foo1234!@#$)
after: nil / nil

Ah, that makes sense. Thanks Greg.

Yeah, immortal strings sound like a good theory. Even this gives the expected result!

let x = "foo"
var foo: NSObject? = x as NSString
weak var fooWeak = foo
print("---------> before foo set to nil: \(foo) / \(fooWeak)")
foo = nil
print("---------> after foo set to nil: \(foo) / \(fooWeak)")

Thanks, Greg. And everyone!