Comment 1 for bug 1078732

Some test is adding non-ascii characters to the test output. There isn't much to go on as the error occurs when the entire xml result file is being generated.

A hack to workaround this is to patch /usr/lib/python2.7/codecs.py StreamWriter.write to do this:
    def write(self, object):

        """ Writes the object's contents encoded to self.stream.
        """
        data, consumed = self.encode(unicode(object, errors=self.errors), self.errors)
        self.stream.write(data)

stream can then be opened with:
stream = open("/tmp/foo.xml", 'w', encoding='utf-8', errors='replace')

This will at least allow the output file to be generated with "??" characters replacing the non-ascii characters.