Issue #14579: Fix error handling bug in the utf-16 decoder.

Patch by Serhiy Storchaka.

Lib/test/test_codecs.py

@@ -495,7 +495,19 @@ class UTF16LETest(ReadTest):
         )
 
     def test_errors(self):
-        self.assertRaises(UnicodeDecodeError, codecs.utf_16_le_decode, "\xff", "strict", True)
+        tests = [
+            (b'\xff', u'\ufffd'),
+            (b'A\x00Z', u'A\ufffd'),
+            (b'A\x00B\x00C\x00D\x00Z', u'ABCD\ufffd'),
+            (b'\x00\xd8', u'\ufffd'),
+            (b'\x00\xd8A', u'\ufffd'),
+            (b'\x00\xd8A\x00', u'\ufffdA'),
+            (b'\x00\xdcA\x00', u'\ufffdA'),
+        ]
+        for raw, expected in tests:
+            self.assertRaises(UnicodeDecodeError, codecs.utf_16_le_decode,
+                              raw, 'strict', True)
+            self.assertEqual(raw.decode('utf-16le', 'replace'), expected)
 
 class UTF16BETest(ReadTest):
     encoding = "utf-16-be"
@@ -516,7 +528,19 @@ class UTF16BETest(ReadTest):
         )
 
     def test_errors(self):
-        self.assertRaises(UnicodeDecodeError, codecs.utf_16_be_decode, "\xff", "strict", True)
+        tests = [
+            (b'\xff', u'\ufffd'),
+            (b'\x00A\xff', u'A\ufffd'),
+            (b'\x00A\x00B\x00C\x00DZ', u'ABCD\ufffd'),
+            (b'\xd8\x00', u'\ufffd'),
+            (b'\xd8\x00\xdc', u'\ufffd'),
+            (b'\xd8\x00\x00A', u'\ufffdA'),
+            (b'\xdc\x00\x00A', u'\ufffdA'),
+        ]
+        for raw, expected in tests:
+            self.assertRaises(UnicodeDecodeError, codecs.utf_16_be_decode,
+                              raw, 'strict', True)
+            self.assertEqual(raw.decode('utf-16be', 'replace'), expected)
 
 class UTF8Test(ReadTest):
     encoding = "utf-8"

Misc/NEWS

@@ -9,6 +9,9 @@ What's New in Python 2.7.4
 Core and Builtins
 -----------------
 
+- Issue #14579: Fix error handling bug in the utf-16 decoder.  Patch by
+  Serhiy Storchaka.
+
 - Issue #15368: An issue that caused bytecode generation to be
   non-deterministic when using randomized hashing (-R) has been fixed.
 

Objects/unicodeobject.c

@@ -2564,7 +2564,7 @@ PyUnicode_DecodeUTF16Stateful(const char *s,
         }
 
         /* UTF-16 code pair: */
-        if (q >= e) {
+        if (e - q < 2) {
             errmsg = "unexpected end of data";
             startinpos = (((const char *)q)-2)-starts;
             endinpos = ((const char *)e)-starts;