Allow #[unsafe(may_dangle)]

[mejrs]

Tracking issue: #159799

This PR allows declaring the attribute itself unsafe, rather than the impl:

#![feature(dropck_eyepatch)]

struct PerhapsLoosely<T>(T);

impl<#[unsafe(may_dangle)] T> Drop for PerhapsLoosely<T> {
    fn drop(&mut self) {
        todo!()
    }
}

It is still possible to use unsafe impl<#[may_dangle] T> Drop, for now. Using both, however, is an error.

r? @JonathanBrouwer
cc @Zoxc

Happy to make a MCP if necessary.

[rustbot]

Some changes occurred in compiler/rustc_passes/src/check_attr.rs

cc @jdonszelmann, @JonathanBrouwer

Some changes occurred in compiler/rustc_hir/src/attrs

cc @jdonszelmann, @JonathanBrouwer

Some changes occurred in compiler/rustc_attr_parsing

cc @jdonszelmann, @JonathanBrouwer

[rustbot]

JonathanBrouwer is currently at their maximum review capacity.
They may take a while to respond.

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[JonathanBrouwer]

I'm leaning towards this not needing an MCP, but not entirely sure

[mejrs]

I'm leaning towards this not needing an MCP, but not entirely sure

Hmm I've been thinking about it

• adding this should be uncontroversial
• removing unsafe impl Drop and migrating everything requires discussion and coordination

So I think it's OK to merge this but do a MCP before merging followups.

[JonathanBrouwer]

@bors r+ rollup
Sounds reasonable

View changes since this review

[rust-bors]

📌 Commit a892808 has been approved by JonathanBrouwer

It is now in the queue for this repository.

🌲 The tree is currently closed for pull requests below priority 1. This pull request will be tested once the tree is reopened.

Reason for tree closure: spurious CI failures

[JonathanBrouwer]

I would actually like to block this PR on the MCP, it doesn't make sense to merge this if the MCP is rejected
@bors r-

[rust-bors]

This pull request was unapproved.

View changes since this unapproval