From 5d076060a5cb4cf1826b2c04dd86234a3770dd80 Mon Sep 17 00:00:00 2001
From: Yarchik <spoko.dev@gmail.com>
Date: Tue, 23 Jun 2026 20:19:09 +0100
Subject: [PATCH] fix: sort dictionary keys bytewise per BEP 3

encode.dict and encode.dictMap sorted keys with Array.prototype.sort(),
which orders by UTF-16 code unit. BEP 3 requires dictionary keys to be
sorted as raw byte strings (bytewise on their UTF-8 encoding). The two
orders diverge for keys containing an astral (non-BMP) code point: its
leading UTF-16 surrogate (0xD800-0xDBFF) sorts before a BMP key in
U+E000-U+FFFF, but its first UTF-8 byte (0xF0+) sorts after that key's
0xEE/0xEF byte.

Because a torrent info-hash is SHA-1 of the bencoded info dictionary, a
dict with such a key produces a non-canonical encoding and a wrong hash.

Sort by the bytes that are actually emitted: the UTF-8 encoding for
string and number keys, the raw bytes for Buffer keys. ASCII/BMP-only
dictionaries are unaffected.
---
 lib/encode.js       | 14 ++++++++++++--
 test/encode.test.js |  8 ++++++++
 2 files changed, 20 insertions(+), 2 deletions(-)

diff --git a/lib/encode.js b/lib/encode.js
index 0667e1d..b210690 100644
--- a/lib/encode.js
+++ b/lib/encode.js
@@ -74,13 +74,23 @@ encode.number = function (buffers, data) {
   }
 }
 
+function compareKeysByBytes (a, b) {
+  const ab = ArrayBuffer.isView(a) ? a : text2arr(String(a))
+  const bb = ArrayBuffer.isView(b) ? b : text2arr(String(b))
+  const length = Math.min(ab.length, bb.length)
+  for (let i = 0; i < length; i++) {
+    if (ab[i] !== bb[i]) return ab[i] - bb[i]
+  }
+  return ab.length - bb.length
+}
+
 encode.dict = function (buffers, data) {
   buffers.push(buffD)
 
   let j = 0
   let k
   // fix for issue #13 - sorted dicts
-  const keys = Object.keys(data).sort()
+  const keys = Object.keys(data).sort(compareKeysByBytes)
   const kl = keys.length
 
   for (; j < kl; j++) {
@@ -96,7 +106,7 @@ encode.dict = function (buffers, data) {
 encode.dictMap = function (buffers, data) {
   buffers.push(buffD)
 
-  const keys = Array.from(data.keys()).sort()
+  const keys = Array.from(data.keys()).sort(compareKeysByBytes)
 
   for (const key of keys) {
     if (data.get(key) == null) continue
diff --git a/test/encode.test.js b/test/encode.test.js
index cb6f3ab..8e9e371 100644
--- a/test/encode.test.js
+++ b/test/encode.test.js
@@ -22,6 +22,14 @@ test('bencode#encode()', function (t) {
     t.equal(Buffer.from(bencode.encode(data)).toString(), 'd7:integeri12345e6:string11:Hello Worlde')
   })
 
+  t.test('should sort dictionary keys bytewise per BEP 3, including astral keys', function (t) {
+    t.plan(1)
+    // '\uE000' (UTF-8 ee 80 80) must sort before '\u{1F600}' (UTF-8 f0 9f 98 80),
+    // even though the astral key's leading UTF-16 surrogate (d83d) sorts first.
+    const data = { '\u{1F600}': 2, '\uE000': 1 }
+    t.equal(Buffer.from(bencode.encode(data)).toString('hex'), '64333aee8080693165343af09f988069326565')
+  })
+
   t.test('should force keys to be strings', function (t) {
     t.plan(1)
     const data = {