I have an S with some intermediaries to help with the curve interpolation. The first couple of intermediary layers worked fine, than I add one more to get this setup:
And then the interpolation does this. (I noticed the Compressed Regular Intermediaries being the “wrong way” around with their width, but that shouldn’t and doesn’t matter when I swap them).
So, what’s going on here, where is it getting that height glitch from?
Two things:
The kinks in the spine are normal VF kinks you need to fix by eliminating one of the three kink conditions. There are forum posts about this.
The big outlines you see are due to the fact that the intermediate layers are missing from the other masters. Add the same width axis intermediates for all other weight values (Black) and the bug will disappear.
Do you need offcurves in the middle of the spine? I bet they make the spine squiggly by adding extra points in TT outlines.
When I add the same brace layers for Compressed Black the same thing happens.
I thought the way brace layers (unlike bracket layers) worked is that you can drop in stray ones as you need, and Glyphs will interpolate to the nearest master. In fact, I have other fonts where I have brace layers only attached to one layer.
Note my masters are:
wght 100 wdth 100
wght 350 wdth 100
wght 900 wdth 100
wght 100 wdth 500
wght 350 wdth 500
wght 900 wdth 500
Moving the brace layers attached to one master only (no idea if that makes any difference) I can observe:
When one axis value (e.g. wdth) is off-master, the other values have to be on-master (e.g. wght 100, 350 or 900). Changing any one brace layer to two off-master values will distort the interpolation.
This interpolates correctly:
This doesn’t (wght 400 off master):
Nor does this (different off master wdth values):
This is not how I remember brace layers to work. Any value or combination of values should work, no? 
Any combination of brace layers work in Glyphs (and static font export), but variable fonts only support dividing the designspace (by adding masters at certain locations) if the resulting divisions still are rectangular with masters at all the corners.
Right, that intuitively makes sense, I was probably misremembering what kind of magic Glyphs does behind the scenes. So in other words: (For it to work as VF) brace layers must not have more than one axis value off-master, but they can be any mix of such brace layers?
So in other words: (For it to work as VF) brace layers must not have more than one axis value off-master, but they can be any mix of such brace layers?
Yea, of course not, just incidental, because I had the intermediary for all widths present…
