Unhelpful errors for unintended use of box (...) ... syntax

[huonw]

fn main() {
    box (1 + 2) * 3;

    // box (1.0 + 2.0).sqrt()
}

with just the first:

boxerror.rs:2:17: 2:20 error: type `<generic integer #2>` cannot be dereferenced
boxerror.rs:2     box (1 + 2) * 3;
                              ^~~
boxerror.rs:2:5: 2:20 error: only the managed heap and exchange heap are currently supported
boxerror.rs:2     box (1 + 2) * 3;
                  ^~~~~~~~~~~~~~~

with the second uncommented it's a parse error:

boxerror.rs:4:20: 4:21 error: unexpected token: `.`
boxerror.rs:4     box (1.0 + 2.0).sqrt()
                                 ^

The (...) is interpreted as the allocator/placement part of the box expression, rather than just part of the value to be boxed. It would be nice to suggest something like "if you wish to use the default box write box () ...;".

(Also, "amusingly", the code as written above hits #14084)

[frewsxcv]

visiting for triage

I don't know what the end-game plan is for the box keyword, so this is probably still a valid issue. Even with opting-in to other box syntax, the error is still there:

/t/triage $ cat foo.rs
#![feature(box_syntax)]

fn main() {
    box (1 + 2) * 3;

    //box (1.0 + 2.0).sqrt()
}
/t/triage $ rustc foo.rs
foo.rs:4:17: 4:20 error: type `_` cannot be dereferenced
foo.rs:4     box (1 + 2) * 3;
                         ^~~
foo.rs:4:5: 4:20 error: only the managed heap and exchange heap are currently supported [E0066]
foo.rs:4     box (1 + 2) * 3;
             ^~~~~~~~~~~~~~~
error: aborting due to 2 previous errors

/t/triage $ cat foo.rs
#![feature(box_syntax)]

fn main() {
    //box (1 + 2) * 3;

    box (1.0 + 2.0).sqrt()
}
/t/triage $ rustc foo.rs
foo.rs:6:20: 6:21 error: expected expression, found `.`
foo.rs:6     box (1.0 + 2.0).sqrt()
                            ^
foo.rs:6:5: 6:20 help: perhaps you meant `box() (foo)` instead?
foo.rs:6     box (1.0 + 2.0).sqrt()
             ^~~~~~~~~~~~~~~
error: aborting due to previous error

Error messages are slightly different though

[pnkfelix]

cc #22181

I think this problem will just go away with migration away from placement-box and towards placement-in.

[huonw]

Box syntax has changed so this is irrelevant, closing.