Open Radar

 
Summary:
When XML data contains elements consisting of certain offending characters like the en dash (–), umlauts or Emoji, an attempt to convert such XML documents into the ISO-8859-15 character set (and others) fails. Worse, in addition the XML document's validity is broken because a closing quotation mark is omitted in the resulting XML.

Steps to Reproduce:
Paste the following code into a macOS playground:

import Cocoa

// Select an offending character, for example a simple – a.k.a. – a.k.a. U+2013 (nothing too exotic, is it?)
// The bug is also triggered by other characters such as Ä or 🤔.
let char = "–"

// Create a XML document which contains the offending character.
let attribute = XMLElement.attribute(withName: "enDash", stringValue: char) as! XMLNode
let element = XMLElement.element(withName: "Demo", children: nil, attributes: [attribute]) as! XMLElement
let doc = XMLDocument.document(withRootElement: element) as! XMLDocument

// This is the statement that causes the bug.
// Some other encodings, f.ex. "UTF-8" or "ISO-2022-KR", do work correctly.
doc.characterEncoding = "ISO-8859-15"

// This object already contains corrupted XML. The next step is only necessary to visualize it.
let data = doc.xmlData

// Prints
//      <Demo enDash="&#34;></Demo>
// and renders the whole XML invalid because of the missing closing quotation mark after the semicolon.
//
// It *should* print the following:
//      <Demo enDash="–"></Demo>
print(String.init(data: data, encoding: String.Encoding.utf8)!)

Expected Results:
Cocoa should be able to correctly translate the whole XML into the ISO-8859-15 charset.

Actual Results:
It fails converting the offending character, which is bad, but additionally breaks the XML's validity by omitting a closing quotation mark:
<Demo enDash="&#34;></Demo>

Version:
Xcode 8.0 beta (8S128d)
OS X 10.11.5 (15F34)

Configuration:
This bug is present in OS X SDKs down to at least version 10.9