I am developing a typeface with Shavian and Roman glyphs. My kerning groups are shared across the two scripts, as there are several letters with similar shapes, and the Latin kerning makes sense for Shavian too.
I had been focussing on the Shavian kerning until today, when I added some latin specific kerning pairs - @f @o -40, as a specific example. The kerning was working beautifully in Glyphs, but not in any of my exported instances.
When I examined the exported instances, I couldn’t find the @f @o or similar rules. But I did notice that ‘o’ was considered to be in the kerning group ‘u+10452’, which is”kick-shavian". In my glyphspackage, that glyph is in fact in the`@o` lhs kerning group.
After banging my head against the wall for a while, on a whim, I removed my custom glyph order parameter. That fixed it!
My custom glyph order puts the Shavian letters at the front. It seems that this reordering upsets the kerning table export algorithm? I’m fine working without it for now, but it would be nice to be able to specify arbitrary glyph orders without messing up exports.