[varLib] warn when using rvrn substitution for glyphs without codepoint?

Given how rvrn is misdesigned to go first instead of last, it can as far as I understand only sensibly be used for glyphs with a Unicode codepoint, i.e. something that is not the result of another substitution. Maybe varLib should log a warning about this when encountering an input glyph without a codepoint and suggest to either use rclt or manually write out the variation substitution chains in the feature file?

[madig]

Already done in a certain PR in a certain secret repo.

[justvanrossum]

While I agree that rvrn is close to useless for real use, I have since learned that its mere existence is due to the specced behavior: without early processing (“before anything else”), there would be no need for rvrn at all, as the remaining behavior can be done via rclt or rlig. (Paraphrasing John Hudson, who explained it to me patiently while I was still foaming with rage 😃 )

That said, maybe indeed such a warning would be good.

(Note that rclt still doesn’t work in most Adobe apps, another thing worthy of some foaming rage.)