On Fri Nov 15 14:46:06 2024 +0000, Nikolay Sivov wrote:
I think the remaining question here is if we want to remove script/lang lookup logic entirely, or instead prioritize. @sgwaki, any idea if it's possible to test that? Testing manually is fine. To be clear, test font should have at least two differing vert features, with second feature being testable with specific script/lang pair, if that is possible to achieve with plain gdi API at all.
I manually tested the vertical features using a specially crafted TrueType file.
The font file displays different circled numbers as vertical substitutes for Japanese punctuation marks (U+3001, U+3002) based on its lookup table.
If the implementation chooses: - the first vert feature, i.e. no script/lang lookup, the app shows ① (U+2460's glyph). - the first script/lang tag (@@@@), it shows ② (U+2461's). - the 'DFLT' script, it shows ③ (U+2462's). - the 'kana' script, it shows ④ (U+2463's).
What we can see is ① with plain gdi API.
See [gsub.txt](/uploads/9d5ef683f937fc0b48f9f7e833fab930/gsub.txt) file in ttx format for details, please.