Update PyUnicode_DecodeUTF8 from RFC 2279 to RFC 3629.

1) #8271: when a byte sequence is invalid, only the start byte and all the
   valid continuation bytes are now replaced by U+FFFD, instead of replacing
   the number of bytes specified by the start byte.
   See http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf (pages 94-95);
2) 5- and 6-bytes-long UTF-8 sequences are now considered invalid (no changes
   in behavior);
3) Add code and tests to reject surrogates (U+D800-U+DFFF) as defined in
   RFC 3629, but leave it commented out since it's not backward compatible;
4) Change the error messages "unexpected code byte" to "invalid start byte"
   and "invalid data" to "invalid continuation byte";
5) Add an extensive set of tests in test_unicode;
6) Fix test_codeccallbacks because it was failing after this change.

Lib/test/test_codeccallbacks.py

@@ -153,28 +153,30 @@ class CodecCallbackTest(unittest.TestCase):
             sout += "\\U%08x" % sys.maxunicode
         self.assertEqual(sin.encode("iso-8859-15", "backslashreplace"), sout)
 
-    def test_decoderelaxedutf8(self):
+    def test_decoding_callbacks(self):
-        # This is the test for a decoding callback handler,
+        # This is a test for a decoding callback handler
-        # that relaxes the UTF-8 minimal encoding restriction.
+        # that allows the decoding of the invalid sequence
-        # A null byte that is encoded as "\xc0\x80" will be
+        # "\xc0\x80" and returns "\x00" instead of raising an error.
-        # decoded as a null byte. All other illegal sequences
+        # All other illegal sequences will be handled strictly.
-        # will be handled strictly.
         def relaxedutf8(exc):
             if not isinstance(exc, UnicodeDecodeError):
                 raise TypeError("don't know how to handle %r" % exc)
-            if exc.object[exc.start:exc.end].startswith("\xc0\x80"):
+            if exc.object[exc.start:exc.start+2] == "\xc0\x80":
                 return (u"\x00", exc.start+2) # retry after two bytes
             else:
                 raise exc
 
-        codecs.register_error(
+        codecs.register_error("test.relaxedutf8", relaxedutf8)
-            "test.relaxedutf8", relaxedutf8)
 
+        # all the "\xc0\x80" will be decoded to "\x00"
         sin = "a\x00b\xc0\x80c\xc3\xbc\xc0\x80\xc0\x80"
         sout = u"a\x00b\x00c\xfc\x00\x00"
         self.assertEqual(sin.decode("utf-8", "test.relaxedutf8"), sout)
+
+        # "\xc0\x81" is not valid and a UnicodeDecodeError will be raised
         sin = "\xc0\x80\xc0\x81"
-        self.assertRaises(UnicodeError, sin.decode, "utf-8", "test.relaxedutf8")
+        self.assertRaises(UnicodeDecodeError, sin.decode,
+                          "utf-8", "test.relaxedutf8")
 
     def test_charmapencode(self):
         # For charmap encodings the replacement string will be

Lib/test/test_unicode.py

@@ -600,6 +600,164 @@ class UnicodeTest(
         # * strict decoding testing for all of the
         #   UTF8_ERROR cases in PyUnicode_DecodeUTF8
 
+    def test_utf8_decode_valid_sequences(self):
+        sequences = [
+            # single byte
+            ('\x00', u'\x00'), ('a', u'a'), ('\x7f', u'\x7f'),
+            # 2 bytes
+            ('\xc2\x80', u'\x80'), ('\xdf\xbf', u'\u07ff'),
+            # 3 bytes
+            ('\xe0\xa0\x80', u'\u0800'), ('\xed\x9f\xbf', u'\ud7ff'),
+            ('\xee\x80\x80', u'\uE000'), ('\xef\xbf\xbf', u'\uffff'),
+            # 4 bytes
+            ('\xF0\x90\x80\x80', u'\U00010000'),
+            ('\xf4\x8f\xbf\xbf', u'\U0010FFFF')
+        ]
+        for seq, res in sequences:
+            self.assertEqual(seq.decode('utf-8'), res)
+
+        for ch in map(unichr, range(0, sys.maxunicode)):
+            self.assertEqual(ch, ch.encode('utf-8').decode('utf-8'))
+
+    def test_utf8_decode_invalid_sequences(self):
+        # continuation bytes in a sequence of 2, 3, or 4 bytes
+        continuation_bytes = map(chr, range(0x80, 0xC0))
+        # start bytes of a 2-byte sequence equivalent to codepoints < 0x7F
+        invalid_2B_seq_start_bytes = map(chr, range(0xC0, 0xC2))
+        # start bytes of a 4-byte sequence equivalent to codepoints > 0x10FFFF
+        invalid_4B_seq_start_bytes = map(chr, range(0xF5, 0xF8))
+        invalid_start_bytes = (
+            continuation_bytes + invalid_2B_seq_start_bytes +
+            invalid_4B_seq_start_bytes + map(chr, range(0xF7, 0x100))
+        )
+
+        for byte in invalid_start_bytes:
+            self.assertRaises(UnicodeDecodeError, byte.decode, 'utf-8')
+
+        for sb in invalid_2B_seq_start_bytes:
+            for cb in continuation_bytes:
+                self.assertRaises(UnicodeDecodeError, (sb+cb).decode, 'utf-8')
+
+        for sb in invalid_4B_seq_start_bytes:
+            for cb1 in continuation_bytes[:3]:
+                for cb3 in continuation_bytes[:3]:
+                    self.assertRaises(UnicodeDecodeError,
+                                      (sb+cb1+'\x80'+cb3).decode, 'utf-8')
+
+        for cb in map(chr, range(0x80, 0xA0)):
+            self.assertRaises(UnicodeDecodeError,
+                              ('\xE0'+cb+'\x80').decode, 'utf-8')
+            self.assertRaises(UnicodeDecodeError,
+                              ('\xE0'+cb+'\xBF').decode, 'utf-8')
+        # XXX: surrogates shouldn't be valid UTF-8!
+        # see http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf
+        # (table 3-7) and http://www.rfc-editor.org/rfc/rfc3629.txt
+        #for cb in map(chr, range(0xA0, 0xC0)):
+            #sys.__stdout__.write('\\xED\\x%02x\\x80\n' % ord(cb))
+            #self.assertRaises(UnicodeDecodeError,
+                              #('\xED'+cb+'\x80').decode, 'utf-8')
+            #self.assertRaises(UnicodeDecodeError,
+                              #('\xED'+cb+'\xBF').decode, 'utf-8')
+        for cb in map(chr, range(0x80, 0x90)):
+            self.assertRaises(UnicodeDecodeError,
+                              ('\xF0'+cb+'\x80\x80').decode, 'utf-8')
+            self.assertRaises(UnicodeDecodeError,
+                              ('\xF0'+cb+'\xBF\xBF').decode, 'utf-8')
+        for cb in map(chr, range(0x90, 0xC0)):
+            self.assertRaises(UnicodeDecodeError,
+                              ('\xF4'+cb+'\x80\x80').decode, 'utf-8')
+            self.assertRaises(UnicodeDecodeError,
+                              ('\xF4'+cb+'\xBF\xBF').decode, 'utf-8')
+
+    def test_issue8271(self):
+        # Issue #8271: when a byte sequence is invalid, only the start byte
+        # and all the valid continuation bytes should be replaced by U+FFFD,
+        # not the number of bytes specified by the start byte.
+        # See http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf (page 95,
+        # table 3-8, Row 2) for more information about the algorithm used.
+        FFFD = u'\ufffd'
+        sequences = [
+            # invalid start bytes
+            ('\x80', FFFD), # continuation byte
+            ('\x80\x80', FFFD*2), # 2 continuation bytes
+            ('\xc0', FFFD),
+            ('\xc0\xc0', FFFD*2),
+            ('\xc1', FFFD),
+            ('\xc1\xc0', FFFD*2),
+            ('\xc0\xc1', FFFD*2),
+            # with start byte of a 2-byte sequence
+            ('\xc2', FFFD), # only the start byte
+            ('\xc2\xc2', FFFD*2), # 2 start bytes
+            ('\xc2\xc2\xc2', FFFD*3), # 2 start bytes
+            ('\xc2\x41', FFFD+'A'), # invalid continuation byte
+            # with start byte of a 3-byte sequence
+            ('\xe1', FFFD), # only the start byte
+            ('\xe1\xe1', FFFD*2), # 2 start bytes
+            ('\xe1\xe1\xe1', FFFD*3), # 3 start bytes
+            ('\xe1\xe1\xe1\xe1', FFFD*4), # 4 start bytes
+            ('\xe1\x80', FFFD), # only 1 continuation byte
+            ('\xe1\x41', FFFD+'A'), # invalid continuation byte
+            ('\xe1\x41\x80', FFFD+'A'+FFFD), # invalid cb followed by valid cb
+            ('\xe1\x41\x41', FFFD+'AA'), # 2 invalid continuation bytes
+            ('\xe1\x80\x41', FFFD+'A'), # only 1 valid continuation byte
+            ('\xe1\x80\xe1\x41', FFFD*2+'A'), # 1 valid and the other invalid
+            ('\xe1\x41\xe1\x80', FFFD+'A'+FFFD), # 1 invalid and the other valid
+            # with start byte of a 4-byte sequence
+            ('\xf1', FFFD), # only the start byte
+            ('\xf1\xf1', FFFD*2), # 2 start bytes
+            ('\xf1\xf1\xf1', FFFD*3), # 3 start bytes
+            ('\xf1\xf1\xf1\xf1', FFFD*4), # 4 start bytes
+            ('\xf1\xf1\xf1\xf1\xf1', FFFD*5), # 5 start bytes
+            ('\xf1\x80', FFFD), # only 1 continuation bytes
+            ('\xf1\x80\x80', FFFD), # only 2 continuation bytes
+            ('\xf1\x80\x41', FFFD+'A'), # 1 valid cb and 1 invalid
+            ('\xf1\x80\x41\x41', FFFD+'AA'), # 1 valid cb and 1 invalid
+            ('\xf1\x80\x80\x41', FFFD+'A'), # 2 valid cb and 1 invalid
+            ('\xf1\x41\x80', FFFD+'A'+FFFD), # 1 invalid cv and 1 valid
+            ('\xf1\x41\x80\x80', FFFD+'A'+FFFD*2), # 1 invalid cb and 2 invalid
+            ('\xf1\x41\x80\x41', FFFD+'A'+FFFD+'A'), # 2 invalid cb and 1 invalid
+            ('\xf1\x41\x41\x80', FFFD+'AA'+FFFD), # 1 valid cb and 1 invalid
+            ('\xf1\x41\xf1\x80', FFFD+'A'+FFFD),
+            ('\xf1\x41\x80\xf1', FFFD+'A'+FFFD*2),
+            ('\xf1\xf1\x80\x41', FFFD*2+'A'),
+            ('\xf1\x41\xf1\xf1', FFFD+'A'+FFFD*2),
+            # with invalid start byte of a 4-byte sequence (rfc2279)
+            ('\xf5', FFFD), # only the start byte
+            ('\xf5\xf5', FFFD*2), # 2 start bytes
+            ('\xf5\x80', FFFD*2), # only 1 continuation byte
+            ('\xf5\x80\x80', FFFD*3), # only 2 continuation byte
+            ('\xf5\x80\x80\x80', FFFD*4), # 3 continuation bytes
+            ('\xf5\x80\x41', FFFD*2+'A'), #  1 valid cb and 1 invalid
+            ('\xf5\x80\x41\xf5', FFFD*2+'A'+FFFD),
+            ('\xf5\x41\x80\x80\x41', FFFD+'A'+FFFD*2+'A'),
+            # with invalid start byte of a 5-byte sequence (rfc2279)
+            ('\xf8', FFFD), # only the start byte
+            ('\xf8\xf8', FFFD*2), # 2 start bytes
+            ('\xf8\x80', FFFD*2), # only one continuation byte
+            ('\xf8\x80\x41', FFFD*2 + 'A'), # 1 valid cb and 1 invalid
+            ('\xf8\x80\x80\x80\x80', FFFD*5), # invalid 5 bytes seq with 5 bytes
+            # with invalid start byte of a 6-byte sequence (rfc2279)
+            ('\xfc', FFFD), # only the start byte
+            ('\xfc\xfc', FFFD*2), # 2 start bytes
+            ('\xfc\x80\x80', FFFD*3), # only 2 continuation bytes
+            ('\xfc\x80\x80\x80\x80\x80', FFFD*6), # 6 continuation bytes
+            # invalid start byte
+            ('\xfe', FFFD),
+            ('\xfe\x80\x80', FFFD*3),
+            # other sequences
+            ('\xf1\x80\x41\x42\x43', u'\ufffd\x41\x42\x43'),
+            ('\xf1\x80\xff\x42\x43', u'\ufffd\ufffd\x42\x43'),
+            ('\xf1\x80\xc2\x81\x43', u'\ufffd\x81\x43'),
+            ('\x61\xF1\x80\x80\xE1\x80\xC2\x62\x80\x63\x80\xBF\x64',
+             u'\x61\uFFFD\uFFFD\uFFFD\x62\uFFFD\x63\uFFFD\uFFFD\x64'),
+        ]
+        for n, (seq, res) in enumerate(sequences):
+            self.assertRaises(UnicodeDecodeError, seq.decode, 'utf-8', 'strict')
+            self.assertEqual(seq.decode('utf-8', 'replace'), res)
+            self.assertEqual((seq+'b').decode('utf-8', 'replace'), res+'b')
+            self.assertEqual(seq.decode('utf-8', 'ignore'),
+                             res.replace(u'\uFFFD', ''))
+
     def test_codecs_idna(self):
         # Test whether trailing dot is preserved
         self.assertEqual(u"www.python.org.".encode("idna"), "www.python.org.")

Objects/unicodeobject.c

@@ -1863,24 +1863,24 @@ encode_char:
 
 static
 char utf8_code_length[256] = {
-    /* Map UTF-8 encoded prefix byte to sequence length.  zero means
+    /* Map UTF-8 encoded prefix byte to sequence length.  Zero means
-       illegal prefix.  see RFC 2279 for details */
+       illegal prefix.  See RFC 3629 for details */
-    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 00-0F */
-    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
-    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
-    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 70-7F */
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 80-8F */
     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* B0-BF */
-    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, /* C0-C1 + C2-CF */
-    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, /* D0-DF */
-    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, /* E0-EF */
-    3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+    4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0  /* F0-F4 + F5-FF */
-    4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6, 0, 0
 };
 
 PyObject *PyUnicode_DecodeUTF8(const char *s,
@@ -1897,6 +1897,7 @@ PyObject *PyUnicode_DecodeUTF8Stateful(const char *s,
 {
     const char *starts = s;
     int n;
+    int k;
     Py_ssize_t startinpos;
     Py_ssize_t endinpos;
     Py_ssize_t outpos;
@@ -1939,7 +1940,9 @@ PyObject *PyUnicode_DecodeUTF8Stateful(const char *s,
             else {
                 errmsg = "unexpected end of data";
                 startinpos = s-starts;
-                endinpos = size;
+                endinpos = startinpos+1;
+                for (k=1; (k < size-startinpos) && ((s[k]&0xC0) == 0x80); k++)
+                    endinpos++;
                 goto utf8Error;
             }
         }
@@ -1947,7 +1950,7 @@ PyObject *PyUnicode_DecodeUTF8Stateful(const char *s,
         switch (n) {
 
         case 0:
-            errmsg = "unexpected code byte";
+            errmsg = "invalid start byte";
             startinpos = s-starts;
             endinpos = startinpos+1;
             goto utf8Error;
@@ -1960,70 +1963,67 @@ PyObject *PyUnicode_DecodeUTF8Stateful(const char *s,
 
         case 2:
             if ((s[1] & 0xc0) != 0x80) {
-                errmsg = "invalid data";
+                errmsg = "invalid continuation byte";
                 startinpos = s-starts;
-                endinpos = startinpos+2;
+                endinpos = startinpos + 1;
                 goto utf8Error;
             }
             ch = ((s[0] & 0x1f) << 6) + (s[1] & 0x3f);
-            if (ch < 0x80) {
+            assert ((ch > 0x007F) && (ch <= 0x07FF));
-                startinpos = s-starts;
+            *p++ = (Py_UNICODE)ch;
-                endinpos = startinpos+2;
-                errmsg = "illegal encoding";
-                goto utf8Error;
-            }
-            else
-                *p++ = (Py_UNICODE)ch;
             break;
 
         case 3:
+            /* XXX: surrogates shouldn't be valid UTF-8!
+               see http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf
+               (table 3-7) and http://www.rfc-editor.org/rfc/rfc3629.txt
+               Uncomment the 2 lines below to make them invalid,
+               codepoints: d800-dfff; UTF-8: \xed\xa0\x80-\xed\xbf\xbf. */
             if ((s[1] & 0xc0) != 0x80 ||
-                (s[2] & 0xc0) != 0x80) {
+                (s[2] & 0xc0) != 0x80 ||
-                errmsg = "invalid data";
+                ((unsigned char)s[0] == 0xE0 &&
+                 (unsigned char)s[1] < 0xA0)/* ||
+                ((unsigned char)s[0] == 0xED &&
+                 (unsigned char)s[1] > 0x9F)*/) {
+                errmsg = "invalid continuation byte";
                 startinpos = s-starts;
-                endinpos = startinpos+3;
+                endinpos = startinpos + 1;
+
+                /* if s[1] first two bits are 1 and 0, then the invalid
+                   continuation byte is s[2], so increment endinpos by 1,
+                   if not, s[1] is invalid and endinpos doesn't need to
+                   be incremented. */
+                if ((s[1] & 0xC0) == 0x80)
+                    endinpos++;
                 goto utf8Error;
             }
             ch = ((s[0] & 0x0f) << 12) + ((s[1] & 0x3f) << 6) + (s[2] & 0x3f);
-            if (ch < 0x0800) {
+            assert ((ch > 0x07FF) && (ch <= 0xFFFF));
-                /* Note: UTF-8 encodings of surrogates are considered
+            *p++ = (Py_UNICODE)ch;
-                   legal UTF-8 sequences;
-
-                   XXX For wide builds (UCS-4) we should probably try
-                   to recombine the surrogates into a single code
-                   unit.
-                */
-                errmsg = "illegal encoding";
-                startinpos = s-starts;
-                endinpos = startinpos+3;
-                goto utf8Error;
-            }
-            else
-                *p++ = (Py_UNICODE)ch;
             break;
 
         case 4:
             if ((s[1] & 0xc0) != 0x80 ||
                 (s[2] & 0xc0) != 0x80 ||
-                (s[3] & 0xc0) != 0x80) {
+                (s[3] & 0xc0) != 0x80 ||
-                errmsg = "invalid data";
+                ((unsigned char)s[0] == 0xF0 &&
+                 (unsigned char)s[1] < 0x90) ||
+                ((unsigned char)s[0] == 0xF4 &&
+                 (unsigned char)s[1] > 0x8F)) {
+                errmsg = "invalid continuation byte";
                 startinpos = s-starts;
-                endinpos = startinpos+4;
+                endinpos = startinpos + 1;
+                if ((s[1] & 0xC0) == 0x80) {
+                    endinpos++;
+                    if ((s[2] & 0xC0) == 0x80)
+                        endinpos++;
+                }
                 goto utf8Error;
             }
             ch = ((s[0] & 0x7) << 18) + ((s[1] & 0x3f) << 12) +
-                ((s[2] & 0x3f) << 6) + (s[3] & 0x3f);
+                 ((s[2] & 0x3f) << 6) + (s[3] & 0x3f);
-            /* validate and convert to UTF-16 */
+            assert ((ch > 0xFFFF) && (ch <= 0x10ffff));
-            if ((ch < 0x10000)        /* minimum value allowed for 4
+
-                                         byte encoding */
-                || (ch > 0x10ffff))   /* maximum value allowed for
-                                         UTF-16 */
-            {
-                errmsg = "illegal encoding";
-                startinpos = s-starts;
-                endinpos = startinpos+4;
-                goto utf8Error;
-            }
 #ifdef Py_UNICODE_WIDE
             *p++ = (Py_UNICODE)ch;
 #else
@@ -2039,13 +2039,6 @@ PyObject *PyUnicode_DecodeUTF8Stateful(const char *s,
             *p++ = (Py_UNICODE)(0xDC00 + (ch & 0x03FF));
 #endif
             break;
-
-        default:
-            /* Other sizes are only needed for UCS-4 */
-            errmsg = "unsupported Unicode code range";
-            startinpos = s-starts;
-            endinpos = startinpos+n;
-            goto utf8Error;
         }
         s += n;
         continue;