Index: Objects/unicodeobject.c
===================================================================
--- Objects/unicodeobject.c	(revision 79542)
+++ Objects/unicodeobject.c	(working copy)
@@ -1896,6 +1896,7 @@
 {
     const char *starts = s;
     int n;
+    int k;
     Py_ssize_t startinpos;
     Py_ssize_t endinpos;
     Py_ssize_t outpos;
@@ -1961,7 +1962,9 @@
             if ((s[1] & 0xc0) != 0x80) {
                 errmsg = "invalid data";
                 startinpos = s-starts;
-                endinpos = startinpos+2;
+                endinpos = startinpos + 1;
+                for (k=1; (k < 2) && ((s[k]&0xC0) == 0x80); k++)
+                    endinpos++;
                 goto utf8Error;
             }
             ch = ((s[0] & 0x1f) << 6) + (s[1] & 0x3f);
@@ -1980,7 +1983,12 @@
                 (s[2] & 0xc0) != 0x80) {
                 errmsg = "invalid data";
                 startinpos = s-starts;
-                endinpos = startinpos+3;
+                endinpos = startinpos + 1;
+                if (((s[0] == 0xE0) && (s[1] < 0xA0)) ||
+                    ((s[0] == 0xED) && (s[1] > 0x9F)))
+                    goto utf8Error;
+                for (k=1; (k < 3) && ((s[k]&0xC0) == 0x80); k++)
+                    endinpos++;
                 goto utf8Error;
             }
             ch = ((s[0] & 0x0f) << 12) + ((s[1] & 0x3f) << 6) + (s[2] & 0x3f);
@@ -2007,7 +2015,12 @@
                 (s[3] & 0xc0) != 0x80) {
                 errmsg = "invalid data";
                 startinpos = s-starts;
-                endinpos = startinpos+4;
+                endinpos = startinpos + 1;
+                if (((s[0] == 0xF0) && (s[1] < 0x90)) ||
+                    ((s[0] == 0xF4) && (s[1] > 0x8F)))
+                    goto utf8Error;
+                for (k=1; (k < 4) && ((s[k]&0xC0) == 0x80); k++)
+                    endinpos++;
                 goto utf8Error;
             }
             ch = ((s[0] & 0x7) << 18) + ((s[1] & 0x3f) << 12) +
Index: Lib/test/test_bytes.py
===================================================================
--- Lib/test/test_bytes.py	(revision 79542)
+++ Lib/test/test_bytes.py	(working copy)
@@ -189,6 +189,55 @@
         self.assertEqual(b.decode(errors="ignore", encoding="utf8"),
                          "Hello world\n")
 
+    def test_utf8_decode_valid_sequences(self):
+        sequences = [
+            ('a', u'a'), # 1 byte seq
+            ('\xc2\xb0', u'\u00b0'), # 2 bytes seq
+            ('\xe9\x99\xaa', u'\u966a'), # 3 bytes seq
+            ('\xf1\x80\x90\xa0', u'\U00040420'), # 4 bytes seq
+        ]
+        for seq, res in sequences:
+            self.assertEqual(seq.decode('utf-8'), res)
+
+    def test_utf8_decode_invalid_sequences(self):
+        #8271
+        sequences = [
+            ('\x80', u'\ufffd'), # continuation byte
+            ('\xc0', u'\ufffd'), # overlong encoding
+            ('\xc2', u'\ufffd'), # 2 bytes seq with only 1 byte
+            ('\xc2\x41', u'\ufffdA'), # 2 bytes seq without continuation byte
+            ('\xe0', u'\ufffd'), # 3 bytes seq with only 1 byte
+            ('\xe0\x80', u'\ufffd'), # 3 bytes seq with only 2 bytes
+            ('\xe0\x81\x41', u'\ufffdA'), # 3 bytes seq with only 2 valid bytes
+            ('\xf0', u'\ufffd'), # 4 bytes seq with only 1 byte
+            ('\xf0\x82', u'\ufffd'), # 4 bytes seq with only 2 bytes
+            ('\xf0\x83\x84', u'\ufffd'), # 4 bytes seq with only 3 bytes
+            ('\xf0\x83\x84\x41', u'\ufffdA'), # 4 bytes seq with only 3 valid bytes
+            ('\xf5', u'\ufffd'), # invalid 4 bytes seq with only 1 byte
+            #('\xf5\x85', u'\ufffd'), # invalid 4 bytes seq with only 2 bytes
+            #('\xf5\x80\x41', u'\ufffdA'), # invalid 4 bytes seq with only 2 bytes
+            #('\xf5\x86\x87\x88', u'\ufffd'), # invalid 4 bytes seq with 4 bytes
+            ('\xf8', u'\ufffd'), # invalid 5 bytes seq with only 1 byte
+            #('\xf8\x89', u'\ufffd'), # invalid 5 bytes seq with only 2 bytes
+            #('\xf8\x80\x41', u'\ufffdA'), # invalid 4 bytes seq with only 2 bytes
+            #('\xf8\x90\x91\x92\x93', u'\ufffd'), # invalid 5 bytes seq with 5 bytes
+            ('\xfc', u'\ufffd'), # invalid 6 bytes seq with only 1 byte
+            #('\xfc\x94\x95', u'\ufffd'), # invalid 6 bytes seq with only 3 bytes
+            #('\xfc\x96\x97\x98\x99\xa0', u'\ufffd'), # invalid 6 bytes seq with 6 bytes
+            ('\xfe', u'\ufffd'), # invalid seq
+            #('\xfe\xa1\xa2', u'\ufffd\ufffd\ufffd'), # invalid seq with 3 bytes
+            # 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):
+            #sys.__stdout__.write('%d %r %r %r\n' % (n, seq, res, seq.decode('utf-8', 'replace')))
+            self.assertRaises(UnicodeDecodeError, seq.decode, 'utf-8')
+            self.assertEqual(seq.decode('utf-8', 'replace'), res)
+
     def test_from_int(self):
         b = self.type2test(0)
         self.assertEqual(b, self.type2test())