On Thu, May 31, 2012 at 01:00:31AM -0400, Behdad Esfahbod wrote:
> On 2012-05-28, at 10:29 PM, Khaled Hosny <khaledho...@eglug.org> wrote:

> > On Mon, May 28, 2012 at 07:50:55PM -0400, Behdad Esfahbod wrote:
> >> [CC'ing Khaled and HarfBuzz list]

> >> This block:

> >> feature calt {
> >>  sub                                              @YearSign @Digits' lookup digit2medium;
> >>  sub @Digits.medium                               @YearSign @Digits' lookup digit2medium;
> >>  sub @Digits.medium @Digits.medium                @YearSign @Digits' lookup digit2medium;
> >>  sub @Digits.medium @Digits.medium @Digits.medium @YearSign @Digits' lookup digit2medium;
> >> } calt;

> >> Looks completely wrong to me in that it matches the backtrack sequence in the
> >> wrong order.  Ie, from what I understand, and I can be totally wrong, it
> >> should have lines like this:

> >>  sub @YearSign @Digits.medium @Digits.medium    @Digits' lookup digit2medium;

> >> instead of this:

> >>  sub @Digits.medium @Digits.medium    @YearSign @Digits' lookup digit2medium;

> > AFAIK the backtrack coverage has to be in the reverse order, if I do as
> > you suggest, only first digit will be matched. This have been tested
> > with FireFox¹ (hb-ng), Pango (hb-old), LibreOffice (ICU) and MS Word
> > (Uniscribe), so I think it is correct² :)

> Reporting back from the Adobe communication thread:

> So, apparently Adobe meant the backtrack sequence to be in logical
> order, not reversed. Combining this with Khaled's observation of the
> font working can only suggest that Fontforge is producing backtrack
> table in reverse. Fixing that will break all fo ts relying on that
> behavior of Fontforge. Khaled is to investigate more to figure out how
> to proceed.