[HarfBuzz] arabic presentation-forms shaping

[Behdad Esfahbod]

On 04/11/2012 09:11 AM, Jonathan Kew wrote:
> 
> 2) I think the ligature code _should_ skip marks, so that a sequence such as
> <lam, fatha, alef> will still find the lam-alef ligature (and leave the fatha
> after it - which loses the distinction between this and <lam, alef, fatha>,
> but it's better than breaking the ligature).

Ok, I'll try to copy the logic from GSUB then.  Ie. using GDEF.  That would
rely on I fixing synthetic GDEF first though.  If we attach lig_id and
component numbers, the fallback mark positioning layer can in fact then try to
position the mark in the right ligature component.

Fixed the rest.

behdad