Re: [ft-devel] light advance rounding

[Alexei Podtelezhnikov]

>> "linearHoriAdvance: [...] Indeed, the value of metrics.horiAdvance
>> that is returned in the glyph slot is normally rounded to integer
>> pixel coordinates (i.e., being a multiple of 64) [...]"
>
>
> 1) advance.x (== metrics.horiAdvance) is purely important for pixel
> placement?
Yes

> 2) Layout engines use "linearHoriAdvance" to compute layouts?
Yes

> 3) How are fractional advance widths realized (like DirectWrite can do
> them)?
This is the only example of subpixel (1/10 of a pixel) positioning I know
https://code.google.com/archive/p/freetype-gl/
So it is possible. Correct use of rsb_delta and lsb_delta limits the
positioning error to half a pixel.

Perhaps you are right that preserving (not rounding) advance is
important for subpixel positioning.