Verify Tibetan stacking

[brawer]

Could the Noto linter verify Tibetan stacking? For example, it could check that a given list of Unicode sequences is composed to one single glyph.

For example, U+0F41 U+0F9A currently does not look so great in Noto Sans Tibetan (rendered by Apple’s CoreText on MacOS X 10.10.3):

Chris Fynn (chris.fynn@gmail.com) wrote on June 26, 2015:

I have a list of Tibetan glyphs in the Chinese GB/T 20524 standard part A with Unicode equivalents. https://sites.google.com/site/chrisfynn2/home/tibetanscriptfonts/standardization/precomposedtibetan-parta

U+F300 to U+F5F8 has the glyphs / character combinations you need for ordinary Tibetan. U+0FF9 to U+F8FF has the most common additional glyphs/ combinations for religious texts / transliteration etc.

Part B of the same standard has several thousand rarely occurring glyphs / combinations. Together this covers just about everything except some very long stacks which occur only in a few Kalacakra texts - and any system is going to have problems rendering those legibly.

[dougfelt]

So it's a safe assumption that stacked forms in Tibetan are ligatures and not positioned glyphs?

[behdad]

For example, U+0F41 U+0F9A currently does not look so great in Noto Sans Tibetan (rendered by Apple’s CoreText on MacOS X 10.10.3):

You should test with HarfBuzz / Uniscribe first.

[ironhouzi]

I'm also experiencing this problem on Arch Linux, using Libre Office 5.1.0.3 (depends on harfbuzz). I am also using the Infinality Freetype patch set, which I suspect might be messing up the stacking for Tibetan. I'm hoping to do some investigation when I get a new laptop.

I also wanted to add that when I select the Noto Sans Tibetan and copy the characters in question into libre office, the font automatically shifts to Noto Sans Devanagari, which renders the stack correctly for some strange reason. When I force the font back to Noto Sans Tibetan, the stack breaks.

EDIT: Also, I'd love to see this stack in an actual text to know if this is just a contrived example, if it's actually Tibetan, or if it's Sanskrit transliteration.

[brawer]

Same with HarfBuzz.

$ wget https://noto-website-2.storage.googleapis.com/pkgs/NotoSansTibetan-unhinted.zip
$ unzip NotoSansTibetan-unhinted.zip
$ hb-view --version
hb-view (HarfBuzz) 1.2.1
Available shapers: ot,fallback
$ hb-view NotoSansTibetan-Regular.ttf ཁྚ --output-format=png >uni0F41_uni0F9A.png

[brawer]

@ironhouzi, the sequence is from Chris Fynn’s list. See the first post on this bug for context.

[behdad]

Ack. Looks to me like bug in the font.