#3479: unichr(2**32) used to return u'\x00'.

The argument was fetched in a long, but PyUnicode_FromOrdinal takes an int.

(why doesn't gcc issue a truncation warning in this case?)

Lib/test/test_builtin.py

@@ -1297,6 +1297,7 @@ class BuiltinTest(unittest.TestCase):
             )
             self.assertRaises(ValueError, unichr, sys.maxunicode+1)
             self.assertRaises(TypeError, unichr)
+            self.assertRaises((OverflowError, ValueError), unichr, 2**32)
 
     # We don't want self in vars(), so these are static methods
 

Misc/NEWS

@@ -12,6 +12,10 @@ What's New in Python 2.6 beta 3?
 Core and Builtins
 -----------------
 
+- Issue #3479: On platforms where sizeof(int) is smaller than sizeof(long)
+  (64bit Unix, for example), unichr() would truncate its argument and return
+  u'\x00' for unichr(2**32). Now it properly raises an OverflowError.
+
 - Apply security patches from Apple.
 
 - Issue #2542: Now that issubclass() may call arbitrary code, ensure that

Python/bltinmodule.c

@@ -394,9 +394,9 @@ Return a string of one character with ordinal i; 0 <= i < 256.");
 static PyObject *
 builtin_unichr(PyObject *self, PyObject *args)
 {
-	long x;
+	int x;
 
-	if (!PyArg_ParseTuple(args, "l:unichr", &x))
+	if (!PyArg_ParseTuple(args, "i:unichr", &x))
 		return NULL;
 
 	return PyUnicode_FromOrdinal(x);