Mispositioned U+1C2C LEPCHA VOWEL SIGN E below U+1C37 LEPCHA SIGN NUKTA

[dscorbett]

Font

NotoSansLepcha-Regular.otf

Where the fonts came from, and when

Site: https://github.com/googlefonts/noto-fonts/blob/4d7396a4fae475975b367e225f2e8f49292d8837/unhinted/otf/NotoSansLepcha/NotoSansLepcha-Regular.otf
Date: 2022-04-16

Font version

Version 2.002

Issue

U+1C2C LEPCHA VOWEL SIGN E should be positioned below U+1C37 LEPCHA SIGN NUKTA, but that only works in this font below single base letters. Below anything else, like U+25CC DOTTED CIRCLE or the cluster <U+1C00 LEPCHA LETTER KA, U+1C24 LEPCHA SUBJOINED LETTER YA>, the vowel sign overlaps the nukta.

Character data

ᰀ᰷ᰤᰬ◌᰷ᰬ
U+1C00 LEPCHA LETTER KA
U+1C37 LEPCHA SIGN NUKTA
U+1C24 LEPCHA SUBJOINED LETTER YA
U+1C2C LEPCHA VOWEL SIGN E
U+25CC DOTTED CIRCLE
U+1C37 LEPCHA SIGN NUKTA
U+1C2C LEPCHA VOWEL SIGN E

Screenshot

[simoncozens]

OK, I've made some progress on this, but this is another of those cross-ligature weirdness issues. Mark-to-mark attachment is working fine for this sequence:

1c00,1c24,1c37,1c2c (ᰀᰤ᰷ᰬ):

But your sequence:

1c00,1c37,1c24,1c2c (ᰀ᰷ᰤᰬ)

(where the two mark glyphs to be mark-to-mark attached are either side of a glyph which takes part in ligation - or perhaps I'm just hyper-aware of this since we've seen this pattern recently) does not attach:

Note that USE does not perform any reordering here and both sequences are accepted without causing dotted circles.

I think we're going to have to go back to the proposal and check which encoding is correct.

[dscorbett]

The encoding with the nukta before the medial consonant is correct. The Unicode Standard, section 13.12 “Lepcha” specifies the syllable order as <consonant or letter a, nukta, medial -ra, medial -ya, dependent vowel, final consonant sign, syllabic modifier>. It also gives the example of <U+1C00 LEPCHA LETTER KA, U+1C37 LEPCHA SIGN NUKTA, U+1C25 LEPCHA SUBJOINED LETTER RA>.

[simoncozens]

I can't find a way to get mark-to-mark working on this kind of cross-ligature mark situation:

ᰀ᰷ᰬᰀ᰷ᰤᰬ

[uni1C00=0+602|uni1C37=0@-272,0+0|uni1C2C=0@-272,-182+0|uni1C00_1C24=3+840|uni1C37=3@-510,0+0|uni1C2C=3@-510,0+0]

So I think I'm just going throw my hands up and make a ligature of the nukta/vowel e.

[dscorbett]

The fix is incomplete. For example, here is <U+1C00, U+1C37, U+1C25, U+1C2C> in version 2.003.

[dscorbett]

The fix is incomplete in version 2.004. Some sequences have have fixed, like <U+1C13, U+1C37, U+1C25, U+1C2C>, because uni1C13_1C25 is now a ligature glyph, but <U+1C00, U+1C37, U+1C25, U+1C2C> is still broken because uni1C00_1C25 is a base glyph. The relevant 'ccmp' lookup only skips ligature glyphs.

[simoncozens]

Huh. I wasn't using a ccmp lookup this time but mark-to-mark positioning!

[simoncozens]

Looks like my previous fix worked purely by chance. The anchors are there but mkmk is not happening; we're getting ligated marks instead.

[dscorbett]

The fix is incomplete in version 2.005. Here is ⟨ᰂ᰷ᰥᰤᰬ⟩ <U+1C02, U+1C37, U+1C25, U+1C24, U+1C2C>:

[simoncozens]

I think I've got this now. The problem is that the cluster order goes below-base vowels, post-base things, and then below-base marks. One approach to this (which the font followed previously) is making the post-base glyphs into marks and filtering them out, but this makes mark attachment (particularly on OS X for some reason) not quite work. So we have to skip over post-base glyphs manually in the code, and there may be several. I don't know how many there can be in Lepcha. I've allowed for three, which allows for ra+ya+a vowel. Lepcha has a fairly simple vowel inventory; I don't think there can be more than one vowel together, so this should be enough.