Now I've done various bits of optimisation, a good chunk of the time of a Noto build is caught up in instantiating UFOs in order to build static TTFs. I'd like to move towards cutting instances from VFs (that's what the front-end does, so it should be acceptable)...
However, this is currently draft because there are some differences between the cut instances and the UFO-instantiated builds, listed below. So we probably need a gftools-cut-instance which fixes up the problems, and use that instead of purely calling hb-subset.
winAscent and winDescent are not updated, neither with fontTools varLib.instancer or hb-subset.
hb-subset does not correct the macStyle and fsSelection bits for bold weights, leading to problems with style linking. (fontTools OK here)
I got multiple .notdef glyphs with hb-subset. (fontTools OK here)
hb-subset does not update the name table (fontTools does).
Now I've done various bits of optimisation, a good chunk of the time of a Noto build is caught up in instantiating UFOs in order to build static TTFs. I'd like to move towards cutting instances from VFs (that's what the front-end does, so it should be acceptable)...
However, this is currently draft because there are some differences between the cut instances and the UFO-instantiated builds, listed below. So we probably need a
gftools-cut-instancewhich fixes up the problems, and use that instead of purely callinghb-subset.winAscentandwinDescentare not updated, neither withfontTools varLib.instancerorhb-subset.hb-subsetdoes not correct the macStyle and fsSelection bits for bold weights, leading to problems with style linking. (fontTools OK here).notdefglyphs withhb-subset. (fontTools OK here)hb-subsetdoes not update the name table (fontTools does).