This is half of SR-15464, that surfaced after SR-15254 was resolved.

Comment by Casper Jeukendrup (JIRA)

I think I have found a significantly simpler way to reproduce this:

@inlinable // This is crucial
func globalFunc() -> Int {
    return localFunc() 

    // Declaration after return statement
    func localFunc() -> Int {
        return 42
    }
}

Or alternatively:

@inlinable
func globalFunc2() -> () -> Int {
    return localFunc

    func localFunc() -> Int {
        return 42
    }
}

Normally, it seems to work well to use a local function before its declaration, but the problem occurs specifically when the local function is declared after the return statement that uses it in an inlinable function.

Should it even be valid to use a local function before its declaration?

Comment by Casper Jeukendrup (JIRA)

I created a pull request with a fix: #40773

Sorry for the late reply!

I was distracted by toggling public and @inlinable, and wasn't able to make the example code simpler. Thanks for further reducing it!

This pitch from a year ago by @slavapestov suggests that it should be valid to use a local function before its declaration. Even if it should be invalid, then the behaviour should be consistent whether or not the outer function is public or @inlinable.

I think it should be valid, and I see that's what your fix is in the PR.

Comment by Casper Jeukendrup (JIRA)

Thanks for your reply! I didn't know about that pitch yet, but indeed I assumed that it should be valid (just like it is for local structs). Indeed, it turns out that my PR is in line with the proposal, so that's good news 🙂

It would be great if someone could start the CI for my PR, because it seems I don't have permission to do so myself.

@jckarter fixed this: #40792

I verified that this is resolved in the 2022-02-25 trunk snapshot.