Dereferencing boxed tuple doesn't compile properly.

[Mokosha]

I think I've encountered a bug, as was described in this stack overflow question. In particular, I would expect the syntax in the following code to be allowed:

#[derive(Clone, PartialEq, Debug)]
enum Foo {
  Base,
  Branch(Box<(Foo, Foo)>)
}

fn do_something(f: Foo) -> Foo {
  match f {
    Foo::Base => Foo::Base,
    Foo::Branch(pair) => {
      let (f1, f2) = *pair;  // ... What do I do here?
      if f2 == Foo::Base { f1 } else { f2 }
   }
  }
}

fn main() {
  let foo = Foo::Base;
  println!("Foo: {:?}", do_something(foo));
}

Instead, I need to do something like

let p = *pair;
let (f1, f2) = p;

Meta

I tried this on the latest nightly release of rustc using the rust playground and it still doesn't work. I'm still somewhat new with rust so let me know if this isn't a bug!

[leoyvens]

#30104 seems related.

[Aatch]

It is related to #30104. Looks like an issue with how moves and pattern matching interact. One issue is that Box<T> is the only pointer type you can actually do this with and we don't want to special case any more than it already is, so this (and the other issue) may have to wait for work on bringing library smart-pointers up to parity with Box<T>.

[gamazeps]

A simpler example could be:

    let a = Box::new(("foo".to_owned(), "bar".to_owned()));
    let (b, c) = *a;

This has to be fixed with:

    let a = Box::new(("foo".to_owned(), "bar".to_owned()));
    let tmp = *a;
    let (b, c) = tmp;

[aidanhs]

Or

let a = Box::new(("foo".to_owned(), "bar".to_owned()));
let (b, c) = {*a};

[mbrubeck]

One of the duplicate issues notes that this behavior was introduced as a consequence of rust-lang/rfcs#130.

[jonhoo]

Is this a dupe of #16223?

[arielb1]

@jonhoo

This isn't strictly defined by rust-lang/rfcs#130, but it's a variant of the same issue. We want a test that this has stable behavior between ast & mir borrowck https://github.com/rust-lang/rust/blob/5d4d09daf2e8e46839647d4e72b1cbefebad6ece/src/test/compile-fail/borrowck-box-insensitivity.rs

[Mokosha]

Sorry -- just to clarify: this is the intended behavior?

[jonhoo]

@arielb1 I'm not sure I follow. Unless I missed it, the linked .rs does not have any test case similar to the case mentioned above. I guess maybe

rust/src/test/compile-fail/borrowck-box-insensitivity.rs

Line 40 in 5d4d09d

let _y = a.y; //~ ERROR use of partially moved

, though that being an error already sounds questionable?