Ok, I'll take a look. Can you send me the font? That said, The HarfBuzz API uses signed integers for advance. So I'm not sure why this is happening. Where do you observe this behavior again?
b On 13-06-13 05:16 PM, Olivier Berten wrote: > Well... Harfbuzz will be the only one behaving that way then... > That font is rendered fine by Adobe, Microsoft, LibreOffice, ... and fontforge > as well as ttx/fontTools gives a negative advance width... I guess this > doesn't push fixing the font ;-) > > Olivier > > > 2013/6/6 Behdad Esfahbod <[email protected] <mailto:[email protected]>> > > Hi Olivier, > > Thanks for the report. I'm hesitant to make HarfBuzz return negative > advance > values as that can confuse downstream users. If this doesn't prove to be > a > major issue, I think we should ignore it for now. > > behdad > > On 13-06-01 05:40 PM, Olivier Berten wrote: > > Hi! > > > > Although it isn't complying with the TrueType/OpenType specifications, > > some fonts have some glyphs with negative advance width. > > > > Advance width (in htmx) is defined as USHORT/uint16 but it seems > > everybody else reads it as SHORT/int16. > > > > In the attached font, glyph "uni17b7" (khmer character ិ) has a > > negative advance width. You can try it with the string និងសមភាព > > > > I wouldn't call this a bug in harfbuzz since it's a font bug but for > > compatibility reasons, it might be useful to change it... > > > > Olivier > > > > > > > > _______________________________________________ > > HarfBuzz mailing list > > [email protected] <mailto:[email protected]> > > http://lists.freedesktop.org/mailman/listinfo/harfbuzz > > > > -- > behdad > http://behdad.org/ > > -- behdad http://behdad.org/ _______________________________________________ HarfBuzz mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/harfbuzz
