U+0C62 and U+0C63 overlap below-base marks

[dscorbett]

Font

NotoSansTelugu-Regular.otf NotoSerifTelugu-Regular.otf

Where the font came from, and when

Site: https://github.com/notofonts/telugu/releases/tag/NotoSerifTelugu-v2.004 Site: https://github.com/notofonts/telugu/releases/tag/NotoSansTelugu-v2.004 Date: 2023-04-08

Font version

Version 2.004

Issue

U+0C62 TELUGU VOWEL SIGN VOCALIC L and U+0C63 TELUGU VOWEL SIGN VOCALIC LL overlap other below-base marks.

Character data

ఘ్తౢఘ్తౣ U+0C18 TELUGU LETTER GHA U+0C4D TELUGU SIGN VIRAMA U+0C24 TELUGU LETTER TA U+0C62 TELUGU VOWEL SIGN VOCALIC L U+0C18 TELUGU LETTER GHA U+0C4D TELUGU SIGN VIRAMA U+0C24 TELUGU LETTER TA U+0C63 TELUGU VOWEL SIGN VOCALIC LL

Screenshots

[simoncozens]

This seems to have been broken for a long time. When I tried to fix it in the Sans by first swapping the order of the l-/ll-vocalic marks and the subscript (so that the vocalic vowel marks could be anchored onto the bottom of the subscripts), I found that other rules then turned the vocalic vowel marks into spacing marks:

Given that Noto Sans is often used in UI contexts where we want to avoid descenders becoming too deep, I consider that actually a reasonable result, so I decided to stop there. For Noto Serif, however, we have less pressure on vertical metrics and more desire for an "authentic" result, so I wanted to anchor the vowel signs underneath the subscript, as (e.g.) Tiro Telugu does. But the subscripts have no anchors on them at all, and I didn't fancy adding hundreds of anchors to support this rule, so I just hacked it with a contextual positioning: