### Fixed

- Unpaired UTF-16 surrogates produce wrong digests and consume the next code unit in the JS path. Unpaired surrogates are now encoded as U+FFFD, matching `TextEncoder` and Node.js `crypto`. #57

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Yi-Cyuan Chen
2026-11-16 16:37:59 +08:00
co-authored by Claude Opus 5.5
parent ee1c332b60
commit 68b5c32442
27 changed files with 273 additions and 111 deletions
+27 -9
View File
@@ -30,7 +30,7 @@
* [js-md5]{@link https://github.com/emn178/js-md5}
*
* @namespace md5
* @version 0.9.2
* @version 0.10.0
* @author Chen, Yi-Cyuan [emn178@gmail.com]
* @copyright Chen, Yi-Cyuan 2014-2026
* @license MIT
@@ -405,7 +405,7 @@
var result = formatMessage(message);
message = result[0];
var isString = result[1];
var code, index = 0, i, length = message.length, blocks = this.blocks;
var code, code2, index = 0, i, length = message.length, blocks = this.blocks;
var buffer8 = this.buffer8;
while (index &lt; length) {
@@ -431,12 +431,18 @@
buffer8[i++] = 0xe0 | (code >>> 12);
buffer8[i++] = 0x80 | ((code >>> 6) &amp; 0x3f);
buffer8[i++] = 0x80 | (code &amp; 0x3f);
} else {
code = 0x10000 + (((code &amp; 0x3ff) &lt;&lt; 10) | (message.charCodeAt(++index) &amp; 0x3ff));
} else if (code &lt; 0xdc00 &amp;&amp; (code2 = message.charCodeAt(index + 1)) >= 0xdc00 &amp;&amp; code2 &lt; 0xe000) {
++index;
code = 0x10000 + (((code &amp; 0x3ff) &lt;&lt; 10) | (code2 &amp; 0x3ff));
buffer8[i++] = 0xf0 | (code >>> 18);
buffer8[i++] = 0x80 | ((code >>> 12) &amp; 0x3f);
buffer8[i++] = 0x80 | ((code >>> 6) &amp; 0x3f);
buffer8[i++] = 0x80 | (code &amp; 0x3f);
} else {
// unpaired surrogate: U+FFFD
buffer8[i++] = 0xef;
buffer8[i++] = 0xbf;
buffer8[i++] = 0xbd;
}
}
} else {
@@ -451,12 +457,18 @@
blocks[i >>> 2] |= (0xe0 | (code >>> 12)) &lt;&lt; SHIFT[i++ &amp; 3];
blocks[i >>> 2] |= (0x80 | ((code >>> 6) &amp; 0x3f)) &lt;&lt; SHIFT[i++ &amp; 3];
blocks[i >>> 2] |= (0x80 | (code &amp; 0x3f)) &lt;&lt; SHIFT[i++ &amp; 3];
} else {
code = 0x10000 + (((code &amp; 0x3ff) &lt;&lt; 10) | (message.charCodeAt(++index) &amp; 0x3ff));
} else if (code &lt; 0xdc00 &amp;&amp; (code2 = message.charCodeAt(index + 1)) >= 0xdc00 &amp;&amp; code2 &lt; 0xe000) {
++index;
code = 0x10000 + (((code &amp; 0x3ff) &lt;&lt; 10) | (code2 &amp; 0x3ff));
blocks[i >>> 2] |= (0xf0 | (code >>> 18)) &lt;&lt; SHIFT[i++ &amp; 3];
blocks[i >>> 2] |= (0x80 | ((code >>> 12) &amp; 0x3f)) &lt;&lt; SHIFT[i++ &amp; 3];
blocks[i >>> 2] |= (0x80 | ((code >>> 6) &amp; 0x3f)) &lt;&lt; SHIFT[i++ &amp; 3];
blocks[i >>> 2] |= (0x80 | (code &amp; 0x3f)) &lt;&lt; SHIFT[i++ &amp; 3];
} else {
// unpaired surrogate: U+FFFD
blocks[i >>> 2] |= 0xef &lt;&lt; SHIFT[i++ &amp; 3];
blocks[i >>> 2] |= 0xbf &lt;&lt; SHIFT[i++ &amp; 3];
blocks[i >>> 2] |= 0xbd &lt;&lt; SHIFT[i++ &amp; 3];
}
}
}
@@ -839,7 +851,7 @@
var i, result = formatMessage(key);
key = result[0];
if (result[1]) {
var bytes = [], length = key.length, index = 0, code;
var bytes = [], length = key.length, index = 0, code, code2;
for (i = 0; i &lt; length; ++i) {
code = key.charCodeAt(i);
if (code &lt; 0x80) {
@@ -851,12 +863,18 @@
bytes[index++] = (0xe0 | (code >>> 12));
bytes[index++] = (0x80 | ((code >>> 6) &amp; 0x3f));
bytes[index++] = (0x80 | (code &amp; 0x3f));
} else {
code = 0x10000 + (((code &amp; 0x3ff) &lt;&lt; 10) | (key.charCodeAt(++i) &amp; 0x3ff));
} else if (code &lt; 0xdc00 &amp;&amp; (code2 = key.charCodeAt(i + 1)) >= 0xdc00 &amp;&amp; code2 &lt; 0xe000) {
++i;
code = 0x10000 + (((code &amp; 0x3ff) &lt;&lt; 10) | (code2 &amp; 0x3ff));
bytes[index++] = (0xf0 | (code >>> 18));
bytes[index++] = (0x80 | ((code >>> 12) &amp; 0x3f));
bytes[index++] = (0x80 | ((code >>> 6) &amp; 0x3f));
bytes[index++] = (0x80 | (code &amp; 0x3f));
} else {
// unpaired surrogate: U+FFFD
bytes[index++] = 0xef;
bytes[index++] = 0xbf;
bytes[index++] = 0xbd;
}
}
key = bytes;