Possible mismatch between ufos and glyphSets for preprocessor filters

[yanone]

I have a proprietary Glyphs VF here that I want to generate with fontmake, so for the time being I'm not linking a source.

It has three main masters. A few glyphs of the Regular master contain intermediate masters at {125} and {155}.

So in total, these are the masters:

• Hairline
• Regular
• Regular {125}
• Regular {155}
• Black

It crashes because the flattenComponent filter can’t find a glyph in the glyphSet. I’ve traced the issue upwards in the call stack until here: https://github.com/googlefonts/ufo2ft/blob/main/Lib/ufo2ft/preProcessor.py#L282

There appears to be a mismatch between the ufo object and the glyphSet. glyphSet correctly contains only the glyphs that exist in that masters, which is all glyphs for the first Regular master, and significantly less for the two subsequent Regular masters representing the intermediate interpolations. However, the ufo object is identical for the three Regular masters. I could be wrong here, but this seems suspicious. Shouldn't there also be three separate ufo objects for the Regular out of the 5 total?

These are the total ufo objects as contained in self.ufos, and you can identify the middle 3 to be identical by its memory address:

[
<ufoLib2.objects.font.Font 'Oodle Sans Hairline' at 0x11168eaf0>,
<ufoLib2.objects.font.Font 'Oodle Sans Regular' at 0x11168e4c0>,
<ufoLib2.objects.font.Font 'Oodle Sans Regular' at 0x11168e4c0>,
<ufoLib2.objects.font.Font 'Oodle Sans Regular' at 0x11168e4c0>,
<ufoLib2.objects.font.Font 'Oodle Sans Black' at 0x119434a60>
]

And then, in the actual filter, an identical filtration is performed on all masters, but the second or third glyphSet of the Regular master doesn’t contain the required glyphs for that operation, so it throws a KeyError.

This is far as I could go, as I can't concentrate anymore at the moment. I can continue to dig but I need guidance. Anyone have an idea what's going on or where the ufo objects get assembled?

[anthrotype]

the master ufos are generated by glyphsLib. Brace layers get translated to additional layers within the same UFO as their linked master, in this case "Regular". So that is by design that you don't see 5 ufos, but only three, of which one contains two additional layers.
Now, masters can be "sparse" in the sense that any glyphs from the default master that are missing from non-default masters "do not participate" (i.e. produce no deltas).
But here's the catch: because of a limitation of UFO whereby component references can only occur within the same layer (see unified-font-object/ufo-spec#49), if you have a composite glyphs in a layer, the latter must also contain the base glyphs for those.
So as workaround for now, you should add the brace trick to all the base glyphs of the composite glyphs that have the brace trick, if that makes sense..
I think ufo2ft is working as intended, if you like this is more a bug report (feature request?) for glyphsLib. Though I am not sure whether/how this could be hanlded there..

[yanone]

Thank you. I got it to work by adding the intermediate layer to just one component glyph, so that was rather straight forward. Only to end up with other interpolation issues (as the decomposeTransformedComponents filter, too, is working on individual masters only).

I understand that a separate VF-first approach is needed. At least Simon has been speaking of this for a while now and I believe he has partial solutions cooking. So I'll close this for now.