I’m am stuck in the following peculiar situation:
I have a MM typeface, composed from two masters (weight – wt0 and wt1). After careful kerning (started four times from scratch) I ended with 6000 kern pairs, which include Latin, Cyrillic, small caps for both, different numeral sets, fractions and etc. – it’s a big typeface. When building the kern feature and then compiling the font everything works just fine. But, if I split the typeface to the desired instances, for fine tuning and polishing, it seems that after updating the kern feature, every font does not compile (excluding the equivalent of wt0) - ending with “[FATAL] GPOS feature 'kern' causes overflow of offset to a subtable”. The peculiar thing is that, if I split the typeface only to the two masters wt0 and wt1, wt0 compiles ok with 6000 pairs. On the other hand wt1 does return “GPOS overflow..” every time, until I reduce the pairs to approximately 4420, which is -1600 pairs short.
Any ideas what could cause this kind of behavior? Thank you in advance!
note: I am using Fontlab 5.04 win. My class kerning is revolving around the idea to have class for every glyph (not shape type) that is part of composite + Cyrillic : A’ Atilte Adireseis…..cyrillic A…etc. Classes are composed from mixed script characters, so I could only kern for differences.