fix(plist): preserve Unicode separators during XML normalization

Use one XML 1.0 line-ending rule in the shared XML boundary and plist source-span owner. Normalize literal CRLF and CR while retaining NEL, line separator and paragraph separator as observed characters. Keep character references, raw scalar locations and prototype-key omission intact.

Validation: 229 XML/plist source tests, 60 adjacent Ghidra/Apktool/SAZ source tests, and 33 compiled CLI/provider tests passed. Full check:ci passed. Real plutil XML and binary fixtures reproduced the baseline loss and passed CLI plus SDK-validated stdio MCP checks with unchanged input digests. Independent source review found no blocker. Binary conversion losses tracked by #5701, #5703 and #5705 remain separate work.

Closes #5702

src/domain/propertyListKeys.test.ts

@@ -315,7 +315,7 @@ it("bounds nested plist wrappers before recursive decoding", () => {
   ).toEqual(["x"]);
 });
 
-it.each(["\n", "\r\n", "\r", "\u0085", "\u2028", "\u2029"])(
+it.each(["\n", "\r\n", "\r"])(
   "preserves unrelated source values around %j line endings",
   (newline) => {
     const text = plist(
@@ -390,3 +390,18 @@ describe("OpenStep plist nesting preflight", () => {
     });
   });
 });
+
+it.each(["\u0085", "\u2028", "\u2029"])(
+  "preserves key identity and omission around inter-element separator %j",
+  (separator) => {
+    const parsed = parseXmlPropertyList(
+      plist(
+        `<dict>${separator}<key>kept${separator}key</key><string>a${separator}b</string>${separator}<key>__proto__</key><string>omitted</string></dict>`,
+      ),
+    );
+    expect(parsed).toEqual({
+      value: { [`kept${separator}key`]: `a${separator}b` },
+      omittedPrototypeKeys: 1,
+    });
+  },
+);

src/domain/propertyListXml.test.ts

@@ -290,6 +290,13 @@ it("binds raw literals to their nodes across normalized line endings and UTF-16
   ]);
 });
 
+it("preserves Unicode line separators in plist string values", () => {
+  const value = "a\u0085b\u2028c\u2029d";
+  expect(
+    parseXmlPropertyList(`<plist><string>${value}</string></plist>`).value,
+  ).toBe(value);
+});
+
 it.each(["\u0085", "\u2028", "\u2029"])(
   "does not normalize unsupported leading whitespace %j into an observed scalar",
   (prefix) => {

src/domain/propertyListXmlSource.ts

@@ -1,15 +1,12 @@
 import type { Document, Node } from "@xmldom/xmldom";
 import { MAX_PLIST_DEPTH } from "./apple/plistValue.js";
 import { err, ok, type Result } from "./result.js";
-import { parseXmlDocument } from "./xmlDocument.js";
+import { parseXmlDocument, XML_LINE_ENDING } from "./xmlDocument.js";
 import {
   decodeXmlPlistData,
   XmlPlistDataError,
 } from "./propertyListXmlData.js";
 
-// Preserve DOM compatibility normalization while retaining raw scalar spans.
-const LINE_ENDING = /\r[\n\u0085]|[\r\n\u0085\u2028\u2029]/gu;
-
 interface ScalarSpan {
   readonly start: number;
   readonly end: number;
@@ -25,14 +22,14 @@ export class XmlPlistSourceError extends TypeError {
 
 /** One source owner binds lexical scalar values to their validated DOM nodes. */
 export const readXmlPlistSource = (text: string) => {
-  const normalized = text.replace(LINE_ENDING, "\n");
+  const normalized = text.replace(XML_LINE_ENDING, "\n");
   const parsed = parseXmlDocument(normalized, {
     doctype: "allow",
     lineEndings: "preserve",
   });
   if (!parsed.ok) throw parsed.error;
   const document = parsed.value;
-  const lineEndings = new RegExp(LINE_ENDING);
+  const lineEndings = new RegExp(XML_LINE_ENDING);
   let line = 1;
   let lineStart = 0;
   // The depth walk visits elements in source order, so location lookup is

src/domain/xmlDocument.test.ts

@@ -73,3 +73,17 @@ it("applies XML line-ending normalization only when the consumer selects it", ()
   expect(parsed.value.documentElement?.getAttribute("attr")).toBe("a b");
   expect(parsed.value.documentElement?.textContent).toBe("\nx");
 });
+
+it("preserves Unicode separators under XML 1.0 normalization", () => {
+  const parsed = parseXmlDocument(
+    '<root attr="a\u0085b\u2028c\u2029d">a\r\u0085b\r\u2028c\r\n&#13;d\u2029</root>',
+    { doctype: "allow", lineEndings: "normalize" },
+  );
+  if (!parsed.ok) throw parsed.error;
+  expect(parsed.value.documentElement?.getAttribute("attr")).toBe(
+    "a\u0085b\u2028c\u2029d",
+  );
+  expect(parsed.value.documentElement?.textContent).toBe(
+    "a\n\u0085b\n\u2028c\n\rd\u2029",
+  );
+});

src/domain/xmlDocument.ts

@@ -1,12 +1,10 @@
-import {
-  DOMParser,
-  normalizeLineEndings,
-  onErrorStopParsing,
-  type Document,
-} from "@xmldom/xmldom";
+import { DOMParser, onErrorStopParsing, type Document } from "@xmldom/xmldom";
 import { SaxesParser } from "saxes";
 import { err, ok, type Result } from "./result.js";
 
+/** XML 1.0 literal line endings; Unicode separators remain source characters. */
+export const XML_LINE_ENDING = /\r\n?|\n/gu;
+
 export class XmlDocumentParseError extends Error {
   readonly _tag = "XmlDocumentParseError";
   constructor(
@@ -48,7 +46,7 @@ export const parseXmlDocument = (
         normalizeLineEndings:
           policy.lineEndings === "preserve"
             ? (text) => text
-            : normalizeLineEndings,
+            : (text) => text.replace(XML_LINE_ENDING, "\n"),
         onError: onErrorStopParsing,
       }).parseFromString(source, "application/xml"),
     );

tests/boundary/cli/nativePlistCli.test.ts

@@ -228,6 +228,7 @@ describe.skipIf(process.platform !== "darwin")(
             "<key>CFBundleExecutable</key><string>App</string>",
             "<key>CFBundleIdentifier</key><string>com.example.types</string>",
             "<key>Blob</key><data>AAEC</data>",
+            "<key>Separators</key><string>a\u0085b\u2028c\u2029d</string>",
             "<key>Built</key><date>2020-01-01T00:00:00Z</date>",
             // Parse the unsigned extended year as a UTC field, without JS string-date inference.
             "<key>Far</key><date>11507-08-17T05:20:00Z</date>",
@@ -261,6 +262,7 @@ describe.skipIf(process.platform !== "darwin")(
             value: {
               CFBundleIdentifier: "com.example.types",
               Blob: { $plist_type: "data", base64: "AAEC" },
+              Separators: "a\u0085b\u2028c\u2029d",
               Built: {
                 $plist_type: "date",
                 iso8601: "2020-01-01T00:00:00.000Z",

tests/boundary/providers/native/plistSourceFormat.test.ts

@@ -113,6 +113,20 @@ it.for<PlistSourceCase>([
     expected: "binary",
     xmlFallback: true,
   },
+  {
+    description: "Apple binary property list\n",
+    expected: "binary",
+    xmlFallback: true,
+    convertedXml:
+      "<plist><dict><key>nel</key><string>a\u0085b</string><key>ls</key><string>a\u2028b</string><key>ps</key><string>a\u2029b</string><key>lf</key><string>a\nb</string><key>payload</key><data>AA==</data></dict></plist>",
+    expectedValue: {
+      nel: "a\u0085b",
+      ls: "a\u2028b",
+      ps: "a\u2029b",
+      lf: "a\nb",
+      payload: { $plist_type: "data", base64: "AA==" },
+    },
+  },
 ])(
   "keeps source format $expected for $description (XML fallback: $xmlFallback)",
   async (