Position of Nukta U+0CBC with some vowels in Noto Serif Kannada

[YusukeShirakihara]

I’d like to report some feedback from a native Kannada speaker hoping to get fixed in Noto Serif Kannada.

It applies to all Noto Sans Kannada fonts:
https://github.com/notofonts/kannada/releases/download/NotoSerifKannada-v2.004/NotoSerifKannada-v2.004.zip

Example: ಕೀ಼ ಕೇ಼ ಕೋ಼ ಜೀ಼ ಜೇ಼ ಜೋ಼ ಫೀ಼ ಫೇ಼ ಫೋ಼ ಸೀ಼ ಸೇ಼ ಸೋ಼
ಕೀ಼
U+0C95 ಕ
U+0CC0 ೀ
U+0CBC ಼

In these cases, Nukta U+0CBC should be directly under the main consonant.

This case may be similar with the another issue that has been closed.
Noto Sans Kannada: issue with nukta and consonant conjugate · Issue #28 · notofonts/kannada (github.com)

[sridatta1]

The order of encoding in the text should be as follows.
Should be <CONSONANT, NUKTA, VOWEL SIGN>
Like
ಕ಼ೀ which is rendered properly

[simoncozens]

@sridatta1, are you sure that is the correct order of encoding? According to L2/03-068:

The NUKTA sign, which modifies a consonant form, is placed immediately after the consonant (after the terminating vowel in case of a dependent vowel appearing after the consonant) in the memory representation and is attached to that consonant in rendering.

So I think the encoding is fine. The problem is that we split the long vowels in short vowel + length mark, and the length mark is a base to which the nukta doesn't anchor. The fix is to make length mark a spacing combining glyph.

[sridatta1]

Oh,
Let me recheck in the core specification

[sridatta1]

From 476 (17 of pdf)

https://www.unicode.org/versions/Unicode15.0.0/ch12.pdf
The nukta is immediately placed after the consonant

[simoncozens]

Oh well, then the font is fine, the input text is wrong. But we should probably still make length mark into a spacing combining mark anyway.