[HarfBuzz] arabic positioning without gpos

[Evan Martin]

On Thu, Jan 14, 2010 at 12:32 PM, Khaled Hosny <khaledhosny at eglug.org> wrote:
> On Thu, Jan 14, 2010 at 11:53:56AM -0800, Evan Martin wrote:
>> This code obviously works for the other users of Harfbuzz, so I must
>> be missing something in my understanding.  (Maybe the glyph advances
>> for Arabic vowels should always be zero?
>
> AFAIK, yes, actually I think this is true for any glyph of combining
> marks class.

I found this function in Pango:

/**
 * pango_ot_buffer_set_zero_width_marks
 * @buffer: a #PangoOTBuffer
 * @zero_width_marks: %TRUE if characters with a mark class should
 *  be forced to zero width.
 *
 * Sets whether characters with a mark class should be forced to zero width.
 * This setting is needed for proper positioning of Arabic accents,
 * but will produce incorrect results with standard OpenType Indic
 * fonts.
 *
 * Since: 1.6
 **/

Pango calls this from the Arabic shaper.  Perhaps such a hack belongs
in Harfbuzz?