Armenian kerning makes punctuation overlap diaeresis

[dscorbett]

Font

NotoSansArmenian-Regular.otf NotoSerifArmenian-Regular.otf

Where the fonts came from, and when

Site: https://github.com/googlefonts/noto-fonts/blob/c129ad2a857a10a429b20300f82bb90262008c17/unhinted/otf/NotoSansArmenian/NotoSansArmenian-Regular.otf Site: https://github.com/googlefonts/noto-fonts/blob/c129ad2a857a10a429b20300f82bb90262008c17/unhinted/otf/NotoSerifArmenian/NotoSerifArmenian-Regular.otf Date: 2022-04-10

Font version

Version 2.005

Issue

Some kerning pairs make the second glyph overlap a diaeresis on the first glyph. They should not overlap.

Character data

ո̈՛օ̈՛ո̈՜օ̈՜ՠ̈՞ա̈՞ո̈՞օ̈՞ U+0578 ARMENIAN SMALL LETTER VO U+0308 COMBINING DIAERESIS U+055B ARMENIAN EMPHASIS MARK U+0585 ARMENIAN SMALL LETTER OH U+0308 COMBINING DIAERESIS U+055B ARMENIAN EMPHASIS MARK U+0578 ARMENIAN SMALL LETTER VO U+0308 COMBINING DIAERESIS U+055C ARMENIAN EXCLAMATION MARK U+0585 ARMENIAN SMALL LETTER OH U+0308 COMBINING DIAERESIS U+055C ARMENIAN EXCLAMATION MARK U+0560 ARMENIAN SMALL LETTER TURNED AYB U+0308 COMBINING DIAERESIS U+055E ARMENIAN QUESTION MARK U+0561 ARMENIAN SMALL LETTER AYB U+0308 COMBINING DIAERESIS U+055E ARMENIAN QUESTION MARK U+0578 ARMENIAN SMALL LETTER VO U+0308 COMBINING DIAERESIS U+055E ARMENIAN QUESTION MARK U+0585 ARMENIAN SMALL LETTER OH U+0308 COMBINING DIAERESIS U+055E ARMENIAN QUESTION MARK

Screenshots

[simoncozens]

This is obviously a bug, but I don't know how these combinations should be handled; should the Armenian marks go above, or to the side, or ...? https://www.unicode.org/L2/L2017/17315-armenian-diaeresis.pdf is not helpful.

[dscorbett]

These punctuation marks only move leftwards. If there is no empty space to move into, they don’t move. They don’t go above uppercase letters or lowercase letters with ascenders. I assume that the same rules apply to these diacriticized letters: the punctuation marks move slightly to the left if there is some space (as in ⟨ա̈՞⟩) or not at all if there isn’t (as in ⟨ո̈՞⟩).

[simoncozens]

Thanks, I understand the requirements now. Basically we need the presence of the diaeresis to block kerning. This is an interesting puzzle for ufo2ft’s kern writer, which always writes an IgnoreMarks lookup flag for kerning pairs.

[simoncozens]

I think that given there’s only a few glyphs with diaeresis attested, I’m going to make them precomposed glyphs. It’s not a neat, general solution, but maybe it doesn’t need to be.

[dscorbett]

To be clear, I don’t think the diaeresis should block all kerning. In some combinations, like ⟨ա̈՛⟩, the kerning is fine.

Composing the glyphs sounds good enough.

[dscorbett]

There is still some overlap in Noto Serif Armenian 2.008: