Fix for bug #438164: %-formatting using Unicode objects.

This patch also does away with an incompatibility between Jython
and CPython.

Lib/test/test_unicode.py

@@ -362,9 +362,6 @@ if sys.platform[:4] != 'java':
 
 verify(u"%(x)s, %(y)s" % {'x':u"abc", 'y':"def"} == u'abc, def')
 try:
-    if sys.platform[:4] != 'java':
-        value = u"%(x)s, %()s" % {'x':u"abc", u''.encode('utf-8'):"def"}
-    else:
     value = u"%(x)s, %()s" % {'x':u"abc", u'':"def"}
 except KeyError:
     print '*** formatting failed for "%s"' % "u'abc, def'"

Objects/unicodeobject.c

@@ -5300,6 +5300,7 @@ PyObject *PyUnicode_Format(PyObject *format,
 				    "incomplete format key");
 		    goto onError;
 		}
+#if 0
 		/* keys are converted to strings using UTF-8 and
 		   then looked up since Python uses strings to hold
 		   variables names etc. in its namespaces and we
@@ -5307,6 +5308,9 @@ PyObject *PyUnicode_Format(PyObject *format,
 		key = PyUnicode_EncodeUTF8(keystart,
 					   keylen,
 					   NULL);
+#else
+		key = PyUnicode_FromUnicode(keystart, keylen);
+#endif
 		if (key == NULL)
 		    goto onError;
 		if (args_owned) {