complain if the codec doesn't return unicode

benjaminp · benjaminp · commit 22d9ee7e177b · 2013-12-28T10:33:58.000-06:00
diff --git a/Lib/test/bad_coding3.py b/Lib/test/bad_coding3.py
@@ -0,0 +1,2 @@
+# coding: string-escape
+\x70\x72\x69\x6e\x74\x20\x32\x2b\x32\x0a
diff --git a/Lib/test/test_pep263.py b/Lib/test/test_pep263.py
@@ -58,6 +58,11 @@ def test_error_message(self):
         with self.assertRaisesRegexp(SyntaxError, 'BOM'):
             compile('\xef\xbb\xbf# -*- coding: fake -*-\n', 'dummy', 'exec')
 
+    def test_non_unicode_codec(self):
+        with self.assertRaisesRegexp(SyntaxError,
+                                     'codec did not return a unicode'):
+            from test import bad_coding3
+
 
 def test_main():
     test_support.run_unittest(PEP263Test)
diff --git a/Misc/NEWS b/Misc/NEWS
@@ -9,6 +9,9 @@ What's New in Python 2.7.7?
 Core and Builtins
 -----------------
 
+- Raise a better error when non-unicode codecs are used for a file's coding
+  cookie.
+
 - Issue #17976: Fixed potential problem with file.write() not detecting IO error
   by inspecting the return value of fwrite().  Based on patches by Jaakko Moisio
   and Victor Stinner.
diff --git a/Parser/tokenizer.c b/Parser/tokenizer.c
@@ -400,6 +400,12 @@ fp_readl(char *s, int size, struct tok_state *tok)
         buf = PyObject_CallObject(tok->decoding_readline, NULL);
         if (buf == NULL)
             return error_ret(tok);
+        if (!PyUnicode_Check(buf)) {
+            Py_DECREF(buf);
+            PyErr_SetString(PyExc_SyntaxError,
+                            "codec did not return a unicode object");
+            return error_ret(tok);
+        }
     } else {
         tok->decoding_buffer = NULL;
         if (PyString_CheckExact(buf))

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	`+# coding: string-escape`
	`2`	`+\x70\x72\x69\x6e\x74\x20\x32\x2b\x32\x0a`