Fix borrowing issues

[SabrinaJewson]

Previously, code like this would fail to compile:

#[context("context")]
async fn borrows(val: &mut u32) -> anyhow::Result<&u32> {
    Ok(&*val)
}

With this error message:

error[E0373]: async block may outlive the current function, but it borrows `val`, which is owned by the current function
 --> src/lib.rs:4:1
  |
4 | #[context("context")]
  | ^^^^^^^^^^^^^^^^^^^^^ may outlive borrowed value `val`
5 | async fn borrows(val: &mut u32) -> anyhow::Result<&u32> {
6 |     Ok(&*val)
  |          --- `val` is borrowed here
  |
note: async block is returned here
 --> src/lib.rs:4:1
  |
4 | #[context("context")]
  | ^^^^^^^^^^^^^^^^^^^^^
  = note: this error originates in an attribute macro (in Nightly builds, run with -Z macro-backtrace for more info)
help: to force the async block to take ownership of `val` (and any other referenced variables), use the `move` keyword
  |
4 | move #[context("context")]
  |

This can be fixed by using an async move {} block instead of an async {} block.

[andrewhickman]

Thanks for the PR!

Looks like the sync version has the same issue. There's a bit of a tradeoff between supporting code like

#[context("failed to parse config at `{}`", path.as_ref().display())]
pub fn parse_config(path: impl AsRef<Path>) -> anyhow::Result<u32> {
    let text = read_to_string(path.as_ref())?;
    Ok(text.parse()?)
}

which fails when we use a move closure/async move block, and your example.

Do you think a way to opt into the move semantics would be reasonable? e.g.

#[context(move, "context")]
async fn borrows(val: &mut u32) -> anyhow::Result<&u32> {

[SabrinaJewson]

Oh, you're right - I didn't consider that case. #[context(move)] seems like a better solution. I can implement it tomorrow in this PR, but feel free to do it now if you want.

[SabrinaJewson]

I have updated the PR to add support for the move option. Additionally, I implemented a commit that should reduce incorrect borrowck errors in synchronous functions by forcing the closure to be FnOnce.

[andrewhickman]

Looks great, thanks! The trick with forcing the closure to be FnOnce is quite cool. I just had a couple comments about variable hygiene, then it should be ready to merge

[andrewhickman]

Thanks for the PR! I'll release version 0.2.0 with these changes