How to extract static reference from Option?

[Virtlink]

I have a static Option<&'static mut Foo> and I want to get the reference out of it. I know this is unsafe, but I can't seem to be able to figure out how to do this. First I tried this:

pub trait Foo { }

static mut _foo: Option<&'static mut Foo> = None;

pub fn get_foo() -> &'static mut Foo {
    unsafe { match _foo {
            Some(x) => x,
            None => panic!(),
    } }
}

The types should match, but I understand I'm trying to move x, which gives me this error:

src/lib.rs:21:15: 21:19 error: cannot move out of static item [E0507]
src/lib.rs:21         match _foo {
                            ^~~~
src/lib.rs:22:18: 22:19 note: attempting to move value to here
src/lib.rs:22             Some(x) => x,
                               ^
src/lib.rs:22:18: 22:19 help: to prevent the move, use `ref x` or `ref mut x`
                        to capture value by reference

So then I tried to use ref mut x to get a mutable reference, followed by *x to get the object back, but that also doesn't work:

pub trait Foo { }

static mut _foo: Option<&'static mut Foo> = None;

pub fn get_foo() -> &'static mut Foo {
    unsafe { match _foo {
            Some(ref mut x) => *x,
            None => panic!(),
    } }
}

src/lib.rs:22:32: 22:34 error: cannot move out of borrowed content [E0507]
src/lib.rs:22             Some(ref mut x) => *x,
                                             ^~

Finally I tried to clone or copy the &'static mut Foo, which also failed.

How do I fix this?

I'm using Rust 1.5.0-nightly (6108e8c 2015-09-28).

---

P.S. While typing this post I figured I try it on the Rust Playground. It fails to compile on nightly, but succeeds on beta and stable? (I'm writing a kernel and need the nightly features.)

[eefriedman]

The following works on both stable and nightly:

use std::ptr;
pub trait Foo { }
static mut _foo: Option<&'static mut Foo> = None;
pub fn get_foo() -> &'static mut Foo {
    unsafe {
        match ptr::read(&_foo) {
            Some(x) => x,
            None => panic!(),
        }
    }
}

The nightly-only error is a consequence of #28321: a borrow of an &'static mut pointer no longer has static lifetime. Not sure this is true.

[eefriedman]

Hmm... it looks like there's a bug here, though: the following, which should be equivalent to your original testcase, works:

pub trait Foo { }

static mut _foo: Option<&'static mut Foo> = None;

pub fn get_foo() -> &'static mut Foo {
    unsafe { match _foo {
            Some(ref mut x) => &mut **x,
            None => panic!(),
    } }
}

fn main() {}

[Virtlink]

@eefriedman Thanks, that works! I'm using your second suggestion now.

I'm curious to know whether this is indeed a bug.

[steveklabnik]

@rust-lang/lang, is this a bug or not?

[eefriedman]

There's definitely something weird going on here, and I don't think it has anything to do with static lifetimes. Another related testcase:

pub struct Foo;
// Accepted on stable and nightly
pub fn get_foo<'a, 'b>(foo: &'a mut Option<&'b mut Foo>) -> &'a mut Foo {
    match foo {
        &mut Some(ref mut x) => *x,
        &mut None => panic!(),
    }
}
// Rejected on nightly, accepted on stable
pub fn get_foo2<'a>(foo: &'a mut Option<&'a mut Foo>) -> &'a mut Foo {
    match foo {
        &mut Some(ref mut x) => *x,
        &mut None => panic!(),
    }
}

[nikomatsakis]

I'm not 100% sure what's going on. It seems to have to do with the coercion rules. I'm not aware that they changed recently, but I'll have to review the PRs.

triage: P-high (regression)

[arielb1]

This works on 1.4 too, but is broken on 3e6d724 - so something recent, but not too recent.

Oddly enough,

pub fn get_foo2<'a>(foo: &'a mut Option<&'a mut Foo>) -> &'a mut Foo {
    match foo {
        &mut Some(ref mut x) => match x { x => *x },
        &mut None => panic!(),
    }
}

works.

[arielb1]

2be0d0a looks like the problem commit

[dotdash]

Hmm... Interesting. I guess we could revert that change while keeping the
test as trans should handle the no-op adjustment by now. I'll try that
tomorrow.
Am 05.10.2015 19:48 schrieb "arielb1" notifications@github.com:

2be0d0a
2be0d0a
looks like the problem commit

—
Reply to this email directly or view it on GitHub
#28839 (comment).

[nikomatsakis]

Ah, yes, that makes sense. I should have thought of that interaction. Sigh.