From 768235d3f16d163c593fba71c89eed6c7f1b61d5 Mon Sep 17 00:00:00 2001
From: Henri Sivonen <hsivonen@hsivonen.fi>
Date: Thu, 18 Jan 2024 16:12:41 +0200
Subject: [PATCH] Fix NFKD for accented digraph followed by accent (#4530)

Fixes https://github.com/unicode-org/icu4x/discussions/4527
---
 CHANGELOG.md                         |  6 +++++-
 components/normalizer/src/lib.rs     |  4 ++--
 components/normalizer/tests/tests.rs | 22 ++++++++++++++++++++++
 3 files changed, 29 insertions(+), 3 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2e21bcc364b..9ba270ae6e1 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -4,11 +4,15 @@
  - [Remove icu_datagen's dep on `fractional`](https://github.com/unicode-org/icu4x/pull/4472)
    - `icu_datagen@1.4.1`
 
+ - Fix normalization of character whose decomposition contains more than one starter and ends with a non-starter followed by a non-starter
+   with a lower Canonical Combining Class than the last character of the decomposition. (https://github.com/unicode-org/icu4x/pull/4530)
+   - `icu_normalizer@1.4.1`
+
 ## icu4x 1.4 (Nov 16, 2023)
 
 - General
   - MSRV is now 1.67
- 
+
 - Components
     - Compiled data updated to CLDR 44 and ICU 74 (https://github.com/unicode-org/icu4x/pull/4245)
     - `icu_calendar`
diff --git a/components/normalizer/src/lib.rs b/components/normalizer/src/lib.rs
index 067b15d933f..9d71e72287a 100644
--- a/components/normalizer/src/lib.rs
+++ b/components/normalizer/src/lib.rs
@@ -637,7 +637,7 @@ where
                 i += 1;
                 // Half-width kana and iota subscript don't occur in the tails
                 // of these multicharacter decompositions.
-                if decomposition_starts_with_non_starter(trie_value) {
+                if !decomposition_starts_with_non_starter(trie_value) {
                     combining_start = i;
                 }
             }
@@ -676,7 +676,7 @@ where
                 i += 1;
                 // Half-width kana and iota subscript don't occur in the tails
                 // of these multicharacter decompositions.
-                if decomposition_starts_with_non_starter(trie_value) {
+                if !decomposition_starts_with_non_starter(trie_value) {
                     combining_start = i;
                 }
             }
diff --git a/components/normalizer/tests/tests.rs b/components/normalizer/tests/tests.rs
index d95486e5e13..f6d1a1e49c5 100644
--- a/components/normalizer/tests/tests.rs
+++ b/components/normalizer/tests/tests.rs
@@ -1308,6 +1308,28 @@ fn test_utf16_basic() {
     );
 }
 
+#[test]
+fn test_accented_digraph() {
+    let normalizer: DecomposingNormalizer = DecomposingNormalizer::new_nfkd();
+    assert_eq!(
+        normalizer.normalize("\u{01C4}\u{0323}"),
+        "DZ\u{0323}\u{030C}"
+    );
+    assert_eq!(
+        normalizer.normalize("DZ\u{030C}\u{0323}"),
+        "DZ\u{0323}\u{030C}"
+    );
+}
+
+#[test]
+fn test_ddd() {
+    let normalizer: DecomposingNormalizer = DecomposingNormalizer::new_nfd();
+    assert_eq!(
+        normalizer.normalize("\u{0DDD}\u{0334}"),
+        "\u{0DD9}\u{0DCF}\u{0334}\u{0DCA}"
+    );
+}
+
 #[test]
 fn test_is_normalized() {
     let nfd: DecomposingNormalizer = DecomposingNormalizer::new_nfd();