normalization is different between unicode builds, so use a new non-BMP char and add normalization test
1 file changed