Broken metrics in interpolated accented characters

The newest version of Cormorant Garamond suffers from some very ugly metrics glitches in the exported fonts. Here’s the string «AÁA» in the TTF export for Regular weight:

This only applies to the interpolated instances, not the masters. Everything looks normal inside glyphs.

I suspect this problem was created when a Marc from Google Fonts removed the «Scale to UPM» parameters and scaled the UPM to 1000. Or maybe it has to do with all the metrics custom parameters in the font info dialog?

In any case, this broken version is currently live on Google Fonts, and it has 80+ million serves per week, so I’m really anxious to get it fixed… :grimacing:

For the record, here’s the GitHub repo:

That is likely. There have been some issues with UPM scaling in recent versions and they are not completely resolved yet.

Best guess now is to revert to an earlier version of the font.

OK, I’ve suggested that now.

I’m wondering how that’s possible, though, when the masters look good in Glyphs. Shouldn’t the error happen at the export stage, then?

Aha, check this out: /Aacute and /Cacute look normal in Cormorant Infant, but /aacute fails:

Looks like only glyphs that (1) are made from components and (2) are switched into place by the Replace Glyphs custom parameter during export are affected! The same glyphs, which are accessible by stylistic sets in the default Cormorant fonts, look correct there. Does that help you fix the bug?

(Also, does that mean our source files might not be corrupted after all?)

One more thing: Should we try to export the fonts with an older version of Glyphs as an emergency measure? Do you know when the bug first appeared? How would we obtain such an older version?

I think I properly fixed this already. I’ll have a look at your files.

Thanks Georg! :smile:

Meanwhile, I’m going to try out version 960 as a workaround.

Both 960 and 939 fail to produce proper exports.

I tried to export the files CormorantRoman.glyphs with the latest version and tested the ‘Medium’ instance. And the result is fine. Can you tell me how you export, what file you are using and what instance your are testing?

Cormorant Medium is fine. The problem arises with the instances that use the Replace Glyphs custom parameter. Try Cormorant Infant Medium and look at /aacute.

BTW, there’s no file at Do I need to know the build number?

can you deactivate the Decompose Glyphs parameter? That fixed this particular issue. What do you need it for?

It’s for the problem I’ve been telling you about for a while now… if I don’t decompose beforehand, Replace Glyphs will produce stealth errors throughout the font. That’s why the Google Fonts version of Cormorant SC didn’t have an /at sign, for instance. / contained /at as a component, so when Replace Glyphs overwrote the latter with the former, it broke.

I suppose as a workaround I could just decompose my whole Glyphs file before exporting (and then trash the copy), but that’s hardly a satisfying solution on the long run.

Is the bug in the Decompose parameter unfixable?

I just decomposed my whole Cormorant.glyphs and removed the Decompose Glyphs custom parameter from the Cormorant Garamond Medium instance. The problem persists in the new export.

This is using [964].

The problem is not with the export. But the file itself is not proper. Scroll through all glyphs in both masters to trigger auto alignment for all layers. You will see a lot of grey cells. Save the file and open it again and it will stay clean.

I have a look at the missing components.

I have fixed the component problem.

And wouldn’t it be better to remove the ‘normal’ lowercase from the SC fonts? You can do that if you put a ‘Remove Glyphs’ Parameter. It is applied after the ‘Rename Glyphs’ so it need to get the .sc names.

1 Like

Hey Georg,

you’re right; after scrolling through all glyphs in the font view, my exports look right again, even if I do use the Decompose custom parameter.

So do you think the «Scale to UPM» command that Marc used while it was buggy messed up the alignments?

I suppose it would indeed be good to remove the promoted glyphs from the spin-offs. I’m just worried I’d break something new if I did that. :grimacing: