U+0F39 TIBETAN MARK TSA -PHRU gets a dotted circle after a vowel sign

[dscorbett]

U+0F39 TIBETAN MARK TSA -PHRU is in USE subclass CMAbv. U+0F71 TIBETAN VOWEL SIGN AA is overridden to Indic_Syllabic_Category=Nukta, so it is in subclass CMBlw. Other vowel signs are in subclasses VAbv and VBlw. Therefore, U+0F39 must precede all vowel signs. However, NFC reorders it after all vowel signs. Therefore, a normalized string like <U+0F40, U+0F71, U+0F39> will get a dotted circle.

HarfBuzz handles this by normalizing U+0F39 to precede all vowel signs, but USE explicitly doesn’t support normalization, so how is USE supposed to handle this?

[xadxura]

Interesting case. @dscorbett, did HarfBuzz get a report of a real text example that is broken by this? It would be good to look at the failing case to consider how best to proceed here.

[dscorbett]

There was no report, since it’s not broken in HarfBuzz. U+0F39 TIBETAN MARK TSA -PHRU is used in Balti, Tibetan transcriptions of Chinese, and contractions. Andrew West’s contraction list contains examples like “གདིུ༹ན་” of U+0F39 with vowel signs. I don’t have examples for Balti and Chinese but I’m sure they must also use vowel signs.