"Moving" out of a static could be a memcpy

[alexcrichton]

This was discussed at the recent work week, but it appears that our notes to not reflect a decision one way or another. I would expect this code to compile:

use std::sync::atomics;

fn main() {
    let a = atomics::INIT_ATOMIC_UINT;
}

Specifically, you can "move" out of any non-mut static, it just turns in to a memcpy. The contents of the static have already been validated to not own anything and have no destructors, so this should be safe.

[flaper87]

Moves out of static items were disallowed as part of 8784d2f

Relaxing this to just mutable static items should be pretty straightforward. I'm tagging the issue as e-easy, I can mentor.

[malu]

I'd be interested to try this. Is it correct that I would either have to add a new option to categorization in middle/mem_categorization.rs or add a boolean field to the cat_static_item option to represent mutability?

Edit: Ok, I guess that is unnecessary given that I can get the mutability through the cmt_ struct. Am I understanding all this correctly?

I should probably add I'm talking about changing the check_is_legal_to_move_from function in middle/borrowck/gather_loans/gather_moves.rs.

[flaper87]

@malu correct, you don't need to add anything to cat_static_item because cmt already has the information you need.

The change consists in adding an arm for cat_static_item and returning true or false based on the mutability.

[malu]

Please see https://github.com/malu/rust/tree/issue-13233 and tell me if I did something wrong or if something is missing.

[flaper87]

@malu it's easier (and better) to comment on pull-requests. Would you mind submitting a PR ? I can provide feedback there. Thanks a lot for working on this.

[nikomatsakis]

cc me

[huonw]

This role is now filled by const. Closing.