make sure we do not break encoded scripts when extracting expected error messages in runtests.py

runtests.py

 #!/usr/bin/python
 
-import os, sys, re, shutil, unittest, doctest
+import os
+import sys
+import re
+import codecs
+import shutil
+import unittest
+import doctest
 
 WITH_CYTHON = True
 
@@ -237,9 +243,10 @@ class CythonCompileTestCase(unittest.TestCase):
 
     def split_source_and_output(self, directory, module, workdir):
         source_file = os.path.join(directory, module) + '.pyx'
-        source_and_output = open(
-            self.find_module_source_file(source_file), 'rU')
-        out = open(os.path.join(workdir, module + '.pyx'), 'w')
+        source_and_output = codecs.open(
+            self.find_module_source_file(source_file), 'rU', 'ISO-8859-1')
+        out = codecs.open(os.path.join(workdir, module + '.pyx'),
+                          'w', 'ISO-8859-1')
         for line in source_and_output:
             last_line = line
             if line.startswith("_ERRORS"):