From fc51e50bc6b180d9fd7e5ceda4fc9b898b2a233e Mon Sep 17 00:00:00 2001
From: Sehoon Shon <sshon@google.com>
Date: Tue, 17 Mar 2026 01:41:19 -0400
Subject: [PATCH] fix(core): handle surrogate pairs in truncateString (#22754)

---
 packages/core/src/utils/textUtils.test.ts | 38 +++++++++++++++++++++++
 packages/core/src/utils/textUtils.ts      | 32 ++++++++++++++++++-
 2 files changed, 69 insertions(+), 1 deletion(-)

diff --git a/packages/core/src/utils/textUtils.test.ts b/packages/core/src/utils/textUtils.test.ts
index 00143b99e3..c1c572a170 100644
--- a/packages/core/src/utils/textUtils.test.ts
+++ b/packages/core/src/utils/textUtils.test.ts
@@ -102,6 +102,44 @@ describe('truncateString', () => {
   it('should handle empty string', () => {
     expect(truncateString('', 5)).toBe('');
   });
+
+  it('should not slice surrogate pairs', () => {
+    const emoji = '😭'; // \uD83D\uDE2D, length 2
+    const str = 'a' + emoji; // length 3
+
+    // We expect 'a' (len 1). Adding the emoji (len 2) would make it 3, exceeding maxLength 2.
+    expect(truncateString(str, 2, '')).toBe('a');
+    expect(truncateString(str, 1, '')).toBe('a');
+    expect(truncateString(emoji, 1, '')).toBe('');
+    expect(truncateString(emoji, 2, '')).toBe(emoji);
+  });
+
+  it('should handle pre-existing dangling high surrogates at the cut point', () => {
+    // \uD83D is a high surrogate without a following low surrogate
+    const str = 'a\uD83Db';
+    // 'a' (1) + '\uD83D' (1) = 2.
+    // BUT our function should strip the dangling surrogate for safety.
+    expect(truncateString(str, 2, '')).toBe('a');
+  });
+
+  it('should handle multi-code-point grapheme clusters like combining marks', () => {
+    // FORCE Decomposed form (NFD) to ensure 'e' + 'accent' are separate code units
+    // This ensures the test behaves the same on Linux and Mac.
+    const combinedChar = 'e\u0301'.normalize('NFD');
+
+    // In NFD, combinedChar.length is 2.
+    const str = 'a' + combinedChar; // 'a' + 'e' + '\u0301' (length 3)
+
+    // Truncating at 2: 'a' (1) + 'e\u0301' (2) = 3. Too long, should stay at 'a'.
+    expect(truncateString(str, 2, '')).toBe('a');
+    expect(truncateString(str, 1, '')).toBe('a');
+
+    // Truncating combinedChar (len 2) at maxLength 1: too long, should be empty.
+    expect(truncateString(combinedChar, 1, '')).toBe('');
+
+    // Truncating combinedChar (len 2) at maxLength 2: fits perfectly.
+    expect(truncateString(combinedChar, 2, '')).toBe(combinedChar);
+  });
 });
 
 describe('safeTemplateReplace', () => {
diff --git a/packages/core/src/utils/textUtils.ts b/packages/core/src/utils/textUtils.ts
index 1066896bc4..8d4cbfa6d5 100644
--- a/packages/core/src/utils/textUtils.ts
+++ b/packages/core/src/utils/textUtils.ts
@@ -80,7 +80,37 @@ export function truncateString(
   if (str.length <= maxLength) {
     return str;
   }
-  return str.slice(0, maxLength) + suffix;
+
+  // This regex matches a "Grapheme Cluster" manually:
+  // 1. A surrogate pair OR a single character...
+  // 2. Followed by any number of "Combining Marks" (\p{M})
+  // 'u' flag is required for Unicode property escapes
+  const graphemeRegex = /(?:[\uD800-\uDBFF][\uDC00-\uDFFF]|.)\p{M}*/gu;
+
+  let truncatedStr = '';
+  let match: RegExpExecArray | null;
+
+  while ((match = graphemeRegex.exec(str)) !== null) {
+    const segment = match[0];
+
+    // If adding the whole cluster (base char + accent) exceeds maxLength, stop.
+    if (truncatedStr.length + segment.length > maxLength) {
+      break;
+    }
+
+    truncatedStr += segment;
+    if (truncatedStr.length >= maxLength) break;
+  }
+
+  // Final safety check for dangling high surrogates
+  if (truncatedStr.length > 0) {
+    const lastCode = truncatedStr.charCodeAt(truncatedStr.length - 1);
+    if (lastCode >= 0xd800 && lastCode <= 0xdbff) {
+      truncatedStr = truncatedStr.slice(0, -1);
+    }
+  }
+
+  return truncatedStr + suffix;
 }
 
 /**