[HarfBuzz] Positive mark AW [was on fallback shaping...]

[Behdad Esfahbod]

On 11/22/10 20:37, Martin Hosken wrote:
> FWIW what happens in graphite is that each mark is positioned relative to its base and the space allocation box (origin to advance width) for the base is updated to be the maximum of the space allocation boxes of the base and the marks (i.e. take the minimum origin and the maximum advance width locations for each positioned mark).

This is wrong.  It's quite conceivable that a combining mark may be wider than
its base.  It should *not* affect the advance width of the base.

> The danger with this approach is that it is very different from other shaping engines which ignore mark advance on markbase and markmark attachment, much to my frustration when trying to implement stuff. A work around has been to use cursive attachment, but I notice that you don't allow cursive attachment if a glyph is a mark.

Right.  I thought about CursivePos too, but as you pointed out, it wouldn't work.

behdad

> 
> HTH,
> Yours,
> Martin