Use subtyping for UnsafeFnPointer coercion, too

[compiler-errors]

I overlooked this in #129059, which changed MIR typechecking to use subtyping for other fn pointer coercions.

Fixes #129285

[rustbot]

r? @cjgillot

rustbot has assigned @cjgillot.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[cjgillot]

I don't feel comfortable to review this.
r? lcnr

[lcnr]

can you flip the args and check whether there's an existing test which goes from fail to pass to make sure we check for that potential unsoundness?

r=me after that

[compiler-errors]

Yes, tests/ui/nll/mir_check_cast_unsafe_fn.rs checks this, which was added 7 years ago:

#![allow(dead_code)]

fn bar<'a>(input: &'a u32, f: fn(&'a u32) -> &'a u32) -> &'static u32 {
    // Here the NLL checker must relate the types in `f` to the types
    // in `g`. These are related via the `UnsafeFnPointer` cast.
    let g: unsafe fn(_) -> _ = f;
    unsafe { g(input) }
    //~^ ERROR lifetime may not live long enough
}

fn main() {}

[compiler-errors]

@bors r=lcnr rollup

[bors]

📌 Commit a97b41f has been approved by lcnr

It is now in the queue for this repository.