I don't think this ever compiled—you would have had to take the pointer you got from withUnsafePointer and convert to point to just a single CChar. @atrick, what's the shortest correct way to do this?

This is a common case. I don't yet have a solution for the error message problem. There's almost no way for users to know what they need to write in the signature of their closure!

This is the correct, working code:

import Foundation
var bytes: (CChar, CChar, CChar, CChar) = (0x61, 0x62, 0x63, 0)
let name: String = withUnsafePointer(to: &bytes) {
  (ptr: UnsafePointer<(CChar, CChar, CChar, CChar)>) -> String in
  return String(utf8String: UnsafeRawPointer(ptr).assumingMemoryBound(to: CChar.self))!
}

The migrator (on trunk) will actually insert the "UnsafeRawPointer(ptr).assumingMemoryBound(to: CChar.self)" for you automatically. The problem is that the migrator can't handle rewriting the closure signature so it never gets to this point. I think a good fix for that would be to give users a useful error message.

As far as making the code cleaner, I'd like to move to this idiom in the very near future:

import Foundation
var bytes: (CChar, CChar, CChar, CChar) = (0x61, 0x62, 0x63, 0)
let name: String = withUnsafeBytes(of: &bytes) {
  return String(utf8String: $0.assumingMemoryBound(to: CChar.self))!
}

Simply not specifying the type of ptr should also work here.

That's right. This works fine in 3.0:

import Foundation
var bytes: (CChar, CChar, CChar, CChar) = (0x61, 0x62, 0x63, 0)
let name: String = withUnsafePointer(to: &bytes) { ptr -> String in
  return String(utf8String: UnsafeRawPointer(ptr).assumingMemoryBound(to: CChar.self))!

In fact, if the 2.2 code were written like this, then the migrator will handle it automatically:

import Foundation
var bytes: (CChar, CChar, CChar, CChar) = (0x61, 0x62, 0x63, 0)
let name: String = withUnsafePointer(&bytes) { ptr -> String in
  return String(UTF8String: UnsafePointer(ptr))!
}

@belkadan This error message is going to cause a lot of grief. Why is it talking about the closure's 'String' return value? And do we already have a bug report for that?

We've been seeing a fair number of these lately: any mismatch in closure arguments and the diagnostics go ridiculous. cc @rudkx

Comment by George King (JIRA)

Thanks for the guidance, and you are correct about the above not being the actual 2.2 code. My apologies! For the record, what I previously had was:

      var d_name = entryPtr?.pointee.d_name # tuple of 256 CChar from a dirent struct.
      var name = ""
      withUnsafePointer(&d_name) {
        name = String(cString: UnsafePointer<Int8>($0))
      }

The workarounds above are sufficient for now.

This use case will be handled gracefully if we get the following feature:

SR-11147: Add withUnsafeBufferPointer(to: variable) where variable is any homogenous aggregate