Minimal change that disables (AFAICT) the interpolation of \u and \U inside

raw string literals.  I added a whole bunch of tests but am still not sure
I am testing all paths through the code.  I really think the code could be
simplified quite a bit.

Lib/test/test_strlit.py

@@ -0,0 +1,148 @@
+r"""Test correct treatment of various string literals by the parser.
+
+There are four types of string literals:
+
+    'abc'   -- normal str
+    r'abc'  -- raw str
+    b'xyz'  -- normal bytes
+    br'xyz' -- raw bytes
+
+The difference between normal and raw strings is of course that in a
+raw string, \ escapes (while still used to determine the end of the
+literal) are not interpreted, so that r'\x00' contains four
+characters: a backslash, an x, and two zeros; while '\x00' contains a
+single character (code point zero).
+
+The tricky thing is what should happen when non-ASCII bytes are used
+inside literals.  For bytes literals, this is considered illegal.  But
+for str literals, those bytes are supposed to be decoded using the
+encoding declared for the file (UTF-8 by default).
+
+We have to test this with various file encodings.  We also test it with
+exec()/eval(), which uses a different code path.
+
+This file is really about correct treatment of encodings and
+backslashes.  It doens't concern itself with issues like single
+vs. double quotes or singly- vs. triply-quoted strings: that's dealt
+with elsewhere (I assume).
+"""
+
+import os
+import sys
+import shutil
+import tempfile
+import unittest
+
+
+TEMPLATE = r"""# coding: %s
+a = 'x'
+assert ord(a) == 120
+b = '\x01'
+assert ord(b) == 1
+c = r'\x01'
+assert list(map(ord, c)) == [92, 120, 48, 49]
+d = '\x81'
+assert ord(d) == 0x81
+e = r'\x81'
+assert list(map(ord, e)) == [92, 120, 56, 49]
+f = '\u1881'
+assert ord(f) == 0x1881
+g = r'\u1881'
+assert list(map(ord, g)) == [92, 117, 49, 56, 56, 49]
+"""
+
+
+def byte(i):
+    return bytes([i])
+
+
+class TestLiterals(unittest.TestCase):
+
+    def setUp(self):
+        self.save_path = sys.path[:]
+        self.tmpdir = tempfile.mkdtemp()
+        sys.path.insert(0, self.tmpdir)
+
+    def tearDown(self):
+        sys.path = self.save_path
+        shutil.rmtree(self.tmpdir, ignore_errors=True)
+
+    def test_template(self):
+        # Check that the template doesn't contain any non-printables
+        # except for \n.
+        for c in TEMPLATE:
+            assert c == '\n' or ' ' <= c <= '~', repr(c)
+
+    def test_eval_str_normal(self):
+        self.assertEqual(eval(""" 'x' """), 'x')
+        self.assertEqual(eval(r""" '\x01' """), chr(1))
+        self.assertEqual(eval(""" '\x01' """), chr(1))
+        self.assertEqual(eval(r""" '\x81' """), chr(0x81))
+        self.assertEqual(eval(""" '\x81' """), chr(0x81))
+        self.assertEqual(eval(r""" '\u1881' """), chr(0x1881))
+        self.assertEqual(eval(""" '\u1881' """), chr(0x1881))
+
+    def test_eval_str_raw(self):
+        self.assertEqual(eval(""" r'x' """), 'x')
+        self.assertEqual(eval(r""" r'\x01' """), '\\' + 'x01')
+        self.assertEqual(eval(""" r'\x01' """), chr(1))
+        self.assertEqual(eval(r""" r'\x81' """), '\\' + 'x81')
+        self.assertEqual(eval(""" r'\x81' """), chr(0x81))
+        self.assertEqual(eval(r""" r'\u1881' """), '\\' + 'u1881')
+        self.assertEqual(eval(""" r'\u1881' """), chr(0x1881))
+
+    def test_eval_bytes_normal(self):
+        self.assertEqual(eval(""" b'x' """), b'x')
+        self.assertEqual(eval(r""" b'\x01' """), byte(1))
+        self.assertEqual(eval(""" b'\x01' """), byte(1))
+        self.assertEqual(eval(r""" b'\x81' """), byte(0x81))
+        self.assertRaises(SyntaxError, eval, """ b'\x81' """)
+        self.assertEqual(eval(r""" b'\u1881' """), b'\\' + b'u1881')
+        self.assertRaises(SyntaxError, eval, """ b'\u1881' """)
+
+    def test_eval_bytes_raw(self):
+        self.assertEqual(eval(""" br'x' """), b'x')
+        self.assertEqual(eval(r""" br'\x01' """), b'\\' + b'x01')
+        self.assertEqual(eval(""" br'\x01' """), byte(1))
+        self.assertEqual(eval(r""" br'\x81' """), b"\\" + b"x81")
+        self.assertRaises(SyntaxError, eval, """ br'\x81' """)
+        self.assertEqual(eval(r""" br'\u1881' """), b"\\" + b"u1881")
+        self.assertRaises(SyntaxError, eval, """ br'\u1881' """)
+
+    def check_encoding(self, encoding, extra=""):
+        modname = "xx_" + encoding.replace("-", "_")
+        fn = os.path.join(self.tmpdir, modname + ".py")
+        f = open(fn, "w", encoding=encoding)
+        try:
+            f.write(TEMPLATE % encoding)
+            f.write(extra)
+        finally:
+            f.close()
+        __import__(modname)
+        del sys.modules[modname]
+
+    def test_file_utf_8(self):
+        extra = "z = '\u1234'; assert ord(z) == 0x1234\n"
+        self.check_encoding("utf-8", extra)
+
+    def test_file_utf_8_error(self):
+        extra = "b'\x80'\n"
+        self.assertRaises(SyntaxError, self.check_encoding, "utf-8", extra)
+
+    def test_file_utf8(self):
+        self.check_encoding("utf8")
+
+    def test_file_iso_8859_1(self):
+        self.check_encoding("iso-8859-1")
+
+    def test_file_latin_1(self):
+        self.check_encoding("latin-1")
+
+    def test_file_latin9(self):
+        self.check_encoding("latin9")
+
+
+if __name__ == "__main__":
+    # Hack so that error messages containing non-ASCII can be printed
+    sys.stdout._encoding = sys.stderr._encoding = "utf-8"
+    unittest.main()

Python/ast.c

@@ -1292,7 +1292,7 @@ ast_for_atom(struct compiling *c, const node *n)
     case STRING: {
         PyObject *str = parsestrplus(c, n, &bytesmode);
         if (!str) {
-            if (PyErr_ExceptionMatches(PyExc_UnicodeError)){
+            if (PyErr_ExceptionMatches(PyExc_UnicodeError)) {
                 PyObject *type, *value, *tback, *errstr;
                 PyErr_Fetch(&type, &value, &tback);
                 errstr = ((PyUnicodeErrorObject *)value)->reason;
@@ -3117,6 +3117,7 @@ decode_unicode(const char *s, size_t len, int rawmode, const char *encoding)
     char *buf;
     char *p;
     const char *end;
+
     if (encoding == NULL) {
         buf = (char *)s;
         u = NULL;
@@ -3218,7 +3219,7 @@ parsestr(const node *n, const char *encoding, int *bytesmode)
             return NULL;
         }
     }
-    if (!*bytesmode) {
+    if (!*bytesmode && !rawmode) {
         return decode_unicode(s, len, rawmode, encoding);
     }
     if (*bytesmode) {
@@ -3238,13 +3239,17 @@ parsestr(const node *n, const char *encoding, int *bytesmode)
     if (rawmode || strchr(s, '\\') == NULL) {
         if (need_encoding) {
             PyObject *v, *u = PyUnicode_DecodeUTF8(s, len, NULL);
-            if (u == NULL)
+            if (u == NULL || !*bytesmode)
-                return NULL;
+                return u;
             v = PyUnicode_AsEncodedString(u, encoding, NULL);
             Py_DECREF(u);
             return v;
-        } else {
+        } else if (*bytesmode) {
             return PyString_FromStringAndSize(s, len);
+        } else if (strcmp(encoding, "utf-8") == 0) {
+            return PyUnicode_FromStringAndSize(s, len);
+	} else {
+            return PyUnicode_DecodeLatin1(s, len, NULL);
         }
     }
 
@@ -3252,7 +3257,7 @@ parsestr(const node *n, const char *encoding, int *bytesmode)
                                  need_encoding ? encoding : NULL);
 }
 
-/* Build a Python string object out of a STRING atom.  This takes care of
+/* Build a Python string object out of a STRING+ atom.  This takes care of
  * compile-time literal catenation, calling parsestr() on each piece, and
  * pasting the intermediate results together.
  */
@@ -3272,8 +3277,7 @@ parsestrplus(struct compiling *c, const node *n, int *bytesmode)
             if (s == NULL)
                 goto onError;
             if (*bytesmode != subbm) {
-                ast_error(n, "cannot mix bytes and nonbytes"
+                ast_error(n, "cannot mix bytes and nonbytes literals");
-                          "literals");
                 goto onError;
             }
             if (PyString_Check(v) && PyString_Check(s)) {