On c.l.py, Martin v. Löwis said that Py_UNICODE could be of a signed type,

so fiddle Jeremy's fix to live with that.  Also added more comments.

Bugfix candidate (this bug is in all versions of Python, at least since
2.1).

Objects/unicodeobject.c

@@ -132,8 +132,12 @@ int unicode_resize(register PyUnicodeObject *unicode,
        instead ! */
     if (unicode == unicode_empty ||
 	(unicode->length == 1 &&
-         /* XXX Is unicode->str[] always unsigned? */
-	 unicode->str[0] < 256U &&
+         /* MvL said unicode->str[] may be signed.  Python generally assumes
+          * an int contains at least 32 bits, and we don't use more than
+          * 32 bits even in a UCS4 build, so casting to unsigned int should
+          * be correct.
+          */
+	 (unsigned int)unicode->str[0] < 256U &&
 	 unicode_latin1[unicode->str[0]] == unicode)) {
         PyErr_SetString(PyExc_SystemError,
                         "can't resize shared unicode objects");
@@ -176,7 +180,7 @@ PyUnicodeObject *_PyUnicode_New(int length)
 {
     register PyUnicodeObject *unicode;
 
-    /* Optimization for empty strings */
+    /* Optimization fo empty strings */
     if (length == 0 && unicode_empty != NULL) {
         Py_INCREF(unicode_empty);
         return unicode_empty;
@@ -213,7 +217,11 @@ PyUnicodeObject *_PyUnicode_New(int length)
 	goto onError;
     }
     /* Initialize the first element to guard against cases where
-       the caller fails before initializing str.
+     * the caller fails before initializing str -- unicode_resize()
+     * reads str[0], and the Keep-Alive optimization can keep memory
+     * allocated for str alive across a call to unicode_dealloc(unicode).
+     * We don't want unicode_resize to read uninitialized memory in
+     * that case.
      */
     unicode->str[0] = 0;
     unicode->str[length] = 0;