Behdad Esfahbod

Compile-time aside, I assumed these all get inlined and simplified. Can we mark the constant ones as such even?

@justvanrossum It's hard to make this in a way that is useful for general use, since the optimizations of which location coordinates and which transformation components to encode also depends...

> Holding off on this until I can determine whether there a consistent byte order used to represent opentype tags as u32. It is looking possible that macOS and Windows...

Are you able to provide fonts to reproduce the issue? CFF is hard to debug for me, but I can give it a try.

According to comments in the merger, merging CID-keyed fonts is not supported. I'm not sure why you don't get the nice error message though.

Fixed the error message... I don't know when / if I can implement this.

The bounds are None for empty glyph indeed. Try: ```diff diff --git a/Lib/fontTools/fontBuilder.py b/Lib/fontTools/fontBuilder.py index dd57a0507..9d06680af 100644 --- a/Lib/fontTools/fontBuilder.py +++ b/Lib/fontTools/fontBuilder.py @@ -117,6 +117,7 @@ charStrings = { } fb.setupCFF(nameStrings["psName"],...

You're not doing anything wrong. The downgrading indeed produces broken fonts sometimes (all the times?!) I'm still scratching my head on it. For me, it produces a font that `ots-sanitize`...

> (One minor problem in [documentation](https://fonttools.readthedocs.io/en/latest/varLib/instancer.html#fontTools.varLib.instancer.instantiateVariableFont): the `varfont` parameter documentation says that CFF2 table is not supported, which is not true anymore as far as I can tell 😄.) Fixed....

> It seems to have to do with the charset. I'll test yours and see if it gives me any clues. Yours also passes ots-sanitize and works with hb-view. But...