Wrong error message with managed boxes

[dhardy]

The compiler reports the wrong error message in this sample (which I don't know whether should be errornous or not; I'm just learning Rust):

fn main() {
    let a = @5;         // immutable

    let mut d = @mut 5; // mutable variable, mutable box
    *d += 5;    // error message about this line, though this is fine
    d = a;      // error doesn't appear if this isn't mentioned

    io::println(fmt!("a, d:\t%d, %d", *a,*d));
}

The message reported is:

dhardy@L10036:~/code/small/rust$ rust run mut.rs 
mut.rs:5:4: 5:6 error: assigning to dereference of const @ pointer
mut.rs:5     *d += 5;
            ^~
error: aborting due to previous error

I'm using the compiler from commit 5f13e9c.

[dhardy]

While we're at it, could someone explain whether "box mutability" is a property of the handle (named variable) or the box? Because:

let mut e = @mut 1;
e = d;  // fine (d from above example)
let mut f = @ 2;
f = d;   // same poor error message as above

[thestinger]

Owned types inherit their mutability from the handle (owner), but managed boxes aren't owned so the mutability is determined by the box's type (@mut vs. @). You could consider the handle to be merely observing it.

@mut and @ are different types, which is why there's an error. The tutorial should be more clear about this. The error message is of course totally wrong :).

[dhardy]

Clear, thanks. So it's a property of both the handle and the box.

Can I be really annoying and ask how to specify that in the type? I.e.

let mut g : @int;

appears to be a mutable handle to an immutable box, but how do I specify a type with a mutable managed box (without assinging it a value)?

[thestinger]

You can use @mut there too, like let x: @mut int;. This also applies to & and &mut since those also just observe a value instead of expressing ownership.

[bstrie]

Reproduced. I presume this would be obviated by moving GC into a library (as it would make it more clear that @ and @mut are separate types).

[alexcrichton]

The error message now accurately states

foo.rs:7:8: 7:9 error: mismatched types: expected `@mut <VI1>` but found `@<VI0>` (values differ in mutability)
foo.rs:7     d = a;      // error doesn't appear if this isn't mentioned
                 ^
error: aborting due to previous error

which is correct because @mut T and @T are different types. Closing this because GC is very soon moving into a library, @-boxes are behind a feature gate, and this seems to be working well enough now (although a test isn't really that necessary with all the in-flight changes)