From 318913baeb6930bf0967ddac9801da5c7053ee05 Mon Sep 17 00:00:00 2001
From: Mihaela Mihaljevic <mihaelamj@me.com>
Date: Tue, 23 Jun 2026 22:23:40 +0200
Subject: [PATCH] fix(emitting): escape CRLF per unicode scalar in
 double-quoted output

A CR+LF pair forms a single Swift extended grapheme cluster, so quote()'s Character-level switch matched neither '\\r' nor '\\n' and emitted the line break unescaped. The folding parser then read the dumped value back differently, breaking parse/dump/parse stability for any string containing CRLF (surfaced as a round-trip failure on the Zoom OpenAPI corpus seed under PUREYAML_RUN_FULL_CORPUS). Iterating over unicode scalars escapes CR and LF independently. Adds a dumper regression test.
---
 CHANGELOG.md                  |  7 +++++++
 Sources/Emitting/Dumper.swift | 10 +++++++---
 Tests/DumperTests.swift       | 18 ++++++++++++++++++
 3 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 71e9564..5dcbf50 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -9,6 +9,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- Escape carriage returns and line feeds in double-quoted output per unicode
+  scalar rather than per Character. A CR+LF pair forms a single Swift extended
+  grapheme cluster, so the previous Character-level escape switch matched neither
+  `\r` nor `\n` and emitted the pair unescaped; the folding parser then read it
+  back differently, breaking parse/dump/parse stability for values containing
+  CRLF (observed on the Zoom OpenAPI corpus seed). Added a dumper regression test.
+
 - Tightened error detection so 15 more official YAML test suite cases are handled
   correctly, repairing four error-detection regressions and removing them from the
   conformance gap manifest. Tabs are now rejected when they would supply
diff --git a/Sources/Emitting/Dumper.swift b/Sources/Emitting/Dumper.swift
index c2fea45..e285c81 100644
--- a/Sources/Emitting/Dumper.swift
+++ b/Sources/Emitting/Dumper.swift
@@ -356,9 +356,13 @@ extension PureYAML.Emitting.Dumper {
     }
 
     func quote(_ value: String) -> String {
+        // Iterate over unicode scalars, not Characters: a CR+LF pair forms a
+        // single extended grapheme cluster in Swift, so a Character-level switch
+        // would match neither "\r" nor "\n" and emit the line break unescaped,
+        // which does not round-trip through the folding parser.
         var output = "\""
-        for character in value {
-            switch character {
+        for scalar in value.unicodeScalars {
+            switch scalar {
             case "\"":
                 output += "\\\""
             case "\\":
@@ -370,7 +374,7 @@ extension PureYAML.Emitting.Dumper {
             case "\t":
                 output += "\\t"
             default:
-                output.append(character)
+                output.unicodeScalars.append(scalar)
             }
         }
         output += "\""
diff --git a/Tests/DumperTests.swift b/Tests/DumperTests.swift
index 876932a..c579035 100644
--- a/Tests/DumperTests.swift
+++ b/Tests/DumperTests.swift
@@ -275,3 +275,21 @@ struct DumperTests {
         #expect(dumper.dump(value) == PureYAML.dump(value))
     }
 }
+
+@Suite("Dumper Line Break Escaping")
+struct DumperLineBreakEscapingTests {
+    @Test("Escapes CR and LF in a CRLF pair so the value round-trips")
+    func test_escapesCarriageReturnLineFeedPair() throws {
+        // A CR+LF pair is a single Swift Character (extended grapheme cluster), so a
+        // Character-level escape switch would emit it unescaped, and the folding
+        // parser would not read it back. Each line break must escape independently.
+        let value = PureYAML.Model.Value.string("a\r\nb\r\n\r\nc")
+        let dumped = PureYAML.dump(value)
+
+        #expect(dumped == "\"a\\r\\nb\\r\\n\\r\\nc\"\n")
+        #expect(!dumped.unicodeScalars.contains("\r"))
+
+        let reparsed = try PureYAML.parse(dumped)
+        #expect(reparsed == value)
+    }
+}