closure/FnMut type mismatch using `box (HEAP) ...` that does not occur with `box ...`

[pnkfelix]

In the program below, using --cfg works1 or --cfg works2 both compile and run successfully, but --cfg breaks produces a type mismatch error.

Not clear on how to fix this or what is actually happening, but it might be blocking the placement overloaded-box desguaring work (rust-lang/rfcs#809)

#![feature(box_syntax)]

use std::cell::RefCell;

type BoxedFunc<'a> = Box<FnMut(&mut u32) -> &u32 + 'a>;
fn to_refcell<'a>(f: BoxedFunc<'a>) -> RefCell<BoxedFunc<'a>>
{
    RefCell::new(f)
}

#[cfg(works1)]
fn make_refcell() -> RefCell<BoxedFunc<'static>>
{
    to_refcell( box () |a:&mut u32| { *a = *a+1; a } )
}


#[cfg(works2)]
fn make_refcell() -> RefCell<BoxedFunc<'static>>
{
    to_refcell( box () |a:&mut u32| { *a = *a+1; a } )
}

#[cfg(breaks)]
fn make_refcell() -> RefCell<BoxedFunc<'static>>
{
    to_refcell( box (::std::boxed::HEAP) |a:&mut u32| { *a = *a+1; a } )
}

fn main() {
    let b = make_refcell();
    let mut a = 9;
    b.into_inner()(&mut a);

    assert_eq!(a, 10);
}

Transcript of error with --cfg breaks, from http://is.gd/NgHq2k :

<anon>:27:17: 27:71 error: type mismatch resolving `for<'r> <[closure <anon>:27:42: 27:71] as core::ops::FnMut<(&'r mut u32,)>>::Output == &'r u32`: values differ in mutability [E0271]
<anon>:27     to_refcell( box (::std::boxed::HEAP) |a:&mut u32| { *a = *a+1; a } )
                          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
<anon>:27:17: 27:71 note: required for the cast to the object type `for<'r> core::ops::FnMut(&'r mut u32) -> &'r u32`
<anon>:27     to_refcell( box (::std::boxed::HEAP) |a:&mut u32| { *a = *a+1; a } )
                          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
error: aborting due to previous error
playpen: application terminated with error code 101

[steveklabnik]

Triage: none of these work today, but I haven't followed the box syntax changes closely enough to know how to update them. Removing the type annotations on the closure arguments leads to

hello.rs:14:29: 14:30 error: unresolved name `a` [E0425]
hello.rs:14         to_refcell( box () |a| { *a = *a+1; a } )
                                        ^

which seems... quite bad.

[Mark-Simulacrum]

Updated code, as best as I can figure. b still breaks with this error. I'm not sure why; I sort of thought that box was just sugar for in (::std::boxed::HEAP), but I guess that's not the case.

error[E0271]: type mismatch resolving `for<'r> <[closure@test.rs:13:31: 13:44] as std::ops::FnOnce<(&'r (),)>>::Output == &'r ()`
  --> test.rs:13:5
   |
13 |     in (::std::boxed::HEAP) { |b:&()| { b } }
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected bound lifetime parameter , found concrete lifetime
   |
   = note: concrete lifetime that was found is lifetime '_#5r
   = note: required for the cast to the object type `for<'r> std::ops::Fn(&'r ()) -> &'r ()`

error: aborting due to previous error

#![feature(box_syntax, box_heap)]
#![feature(placement_in_syntax)]

type BoxedFunc = Box<Fn(&()) -> &()>;

// works
fn a() -> BoxedFunc {
    box |a:&()| { a }
}

// breaks
fn b() -> BoxedFunc {
    in (::std::boxed::HEAP) { |b:&()| { b } }
}

fn main() {}

[steveklabnik]

Triage: as far as I know, everything about box syntax is going back to first base, right? I'm not sure if this issue is useful or not.

[bjorn3]

Placement in syntax has been removed so this issue is no longer relevant I think.