April 10, 2026
5:41 p.m.
In general this looks good. The issue is that we are guessing how many glyphs will result from the string and we where guessing a maximum of 1.5 glyphs per code point and you bump it to 2. This will break if we find a substitution rule that results in more than 2 glyphs per code point and you generate a string of those unicode characters long enough. A quick web search does not find a lot of examples of 3+ glyphs per code point so we are likely safe with this. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10612#note_135829