Closed anthrotype closed 3 weeks ago
turns out this was because we weren't in fact invoking fontc for building static TTFs like this, but fontmake.. should be fixed now with https://github.com/googlefonts/gftools/commit/f14c7bc4e5e17b3262297c1f64d63a0ac717d4c0
we have code in ttx_diff.py to make sure that the order of LookuListIndex values in a Feature record is sorted in the fonts generated by fontmake, because fontc supposedly always has them sorted (and the order doesn't matter anyway):
https://github.com/googlefonts/fontc/blob/0a5482f9366d6eef296158ff4fa2163055593bab/resources/scripts/ttx_diff.py#L408-L421
however, I noticed that when building https://github.com/JAMO-TYPEFACE/Moirai#Sources/Moirai.glyphs in both default and gftools mode, the GPOS tables that fontc (ignore fontmake) generates differ between each other like so:
This is weird. Even weirder is that the only diff when comparing fontc vs fontmake fonts build with
--compare gftools
mode, is exactly the same as above:The GPOS table built by fontmake does not change whether it was built in default or gftools mode.
ttx_diff only applies the sorting of the lookup indices to the font generated by fontmake (be in default or gftools mode), never to the font generated by fontc because the assumption was that those should already be sorted.
But they aren't somehow, and only when running fontc through gftools... 🤔