Problem found while converting from PyBytes to PyString:

Re-enable (and correct) a test for the BOM at the beginning of a code unit.
And properly "unget" characters when the BOM is incomplete.

2 files changed, 22 insertions, 11 deletions

diff --git a/Lib/test/test_builtin.py b/Lib/test/test_builtin.py
index 4f84328..d543751 100644
--- a/Lib/test/test_builtin.py
+++ b/Lib/test/test_builtin.py
@@ -200,8 +200,8 @@ class BuiltinTest(unittest.TestCase):
 
     def test_compile(self):
         compile('print(1)\n', '', 'exec')
-##         bom = b'\xef\xbb\xbf'
-##         compile(bom + b'print(1)\n', '', 'exec')
+        bom = b'\xef\xbb\xbf'
+        compile(bom + b'print(1)\n', '', 'exec')
         compile(source='pass', filename='?', mode='exec')
         compile(dont_inherit=0, filename='tmp', source='0', mode='eval')
         compile('pass', '?', dont_inherit=1, mode='exec')
@@ -327,11 +327,12 @@ class BuiltinTest(unittest.TestCase):
         self.assertEqual(eval('c', globals, locals), 300)
         globals = {'a': 1, 'b': 2}
         locals = {'b': 200, 'c': 300}
-##         bom = b'\xef\xbb\xbf'
-##         self.assertEqual(eval(bom + b'a', globals, locals), 1)
+        bom = b'\xef\xbb\xbf'
+        self.assertEqual(eval(bom + b'a', globals, locals), 1)
         self.assertEqual(eval('"\xe5"', globals), "\xe5")
         self.assertRaises(TypeError, eval)
         self.assertRaises(TypeError, eval, ())
+        self.assertRaises(SyntaxError, eval, bom[:2] + b'a')
 
     def test_general_eval(self):
         # Tests that general mappings can be used for the locals argument
diff --git a/Parser/tokenizer.c b/Parser/tokenizer.c
index 1c2b8e8..2c7da7c 100644
--- a/Parser/tokenizer.c
+++ b/Parser/tokenizer.c
@@ -323,8 +323,21 @@ check_bom(int get_char(struct tok_state *),
 	if (ch == EOF) {
 		return 1;
 	} else if (ch == 0xEF) {
-		ch = get_char(tok); if (ch != 0xBB) goto NON_BOM;
-		ch = get_char(tok); if (ch != 0xBF) goto NON_BOM;
+		ch = get_char(tok); 
+		if (ch != 0xBB) {
+			unget_char(ch, tok);
+			unget_char(0xEF, tok);
+			/* any token beginning with '\xEF' is a bad token */
+			return 1;
+		}
+		ch = get_char(tok); 
+		if (ch != 0xBF) {
+			unget_char(ch, tok);
+			unget_char(0xBB, tok);
+			unget_char(0xEF, tok);
+			/* any token beginning with '\xEF' is a bad token */
+			return 1;
+		}
 #if 0
 	/* Disable support for UTF-16 BOMs until a decision
 	   is made whether this needs to be supported.  */
@@ -344,10 +357,7 @@ check_bom(int get_char(struct tok_state *),
 	if (tok->encoding != NULL)
 		PyMem_FREE(tok->encoding);
 	tok->encoding = new_string("utf-8", 5);	/* resulting is in utf-8 */
-	return 1;
-  NON_BOM:
-	/* any token beginning with '\xEF', '\xFE', '\xFF' is a bad token */
-	unget_char(0xFF, tok);	/* XXX this will cause a syntax error */
+	/* No need to set_readline: input is already utf-8 */
 	return 1;
 }
 
@@ -641,7 +651,7 @@ decode_str(const char *str, struct tok_state *tok)
 		utf8 = translate_into_utf8(str, tok->enc);
 		if (utf8 == NULL)
 			return error_ret(tok);
-		str = PyBytes_AsString(utf8);
+		str = PyString_AsString(utf8);
 	}
 	for (s = str;; s++) {
 		if (*s == '\0') break;