It seems like the normalization tests fail, I'm not quite sure how those work but I'd assume this suggests that changes to theunicode.py script are necessary?

either that, or to the code. Unclear!

This seems to be due tohttps://www.unicode.org/reports/tr15/tr15-56.html#Contexts_Care.

I have no interest in unicode normalization, so I don't plan to look into this. The patch works just fine for everything else.