[HarfBuzz] Post Base Substitution fails with Malayalam repha

[Behdad Esfahbod]

On 13-10-15 08:23 PM, Rajeesh K Nambiar wrote:
> On Tue, Oct 15, 2013 at 6:12 PM, Behdad Esfahbod <behdad at behdad.org> wrote:
>>>
>>> The word 'ചൎച്ച്' <Cha+Dotreph+Cha+Virama+Cha+Virama> is incorrectly
>>> shaped - the final Virama is reordered before
>>> <Dotreph+Cha+Virama+Cha>.
>>> hb-shape output: [ch1=0+1488|xx=1+184|dotrephch1ch1=1+1488]
>>> Correct output: [ch1,dotrephch1ch1,xx].
>>
>> This will be fixed in tonight's push.
> 
> Just pulled the changes and tested - indeed this is fixed!
> 
> We still have an issue with reordering, though :-)
> For the newer versions of Meera and Rachana [¹], we have moved away
> from individual glyphs involving dotreph (it is a lot of glyph
> duplication and causes increased font size) into GPOS "abvm" lookup.
> With this version of fonts, dotreph is still reordered to the end of
> the cluster.

Of course it's still reordered.  That's how "abvm" works: mark should be after
the base for mark attachments to work.


> This causes the final Matra (u2) to be detached which is not desirable
> for traditional orthography.

Same advice: add dotreph to the mark skipping of your ligating table.

behdad


> Ideally, glyph th1th1u2 should have been
> formed with dotreph above it.
> 
> New version of fonts [¹]:
> http://download.savannah.gnu.org/releases/smc/fonts/malayalam-fonts-6.0/Rachana/Rachana.ttf
> , http://download.savannah.gnu.org/releases/smc/fonts/malayalam-fonts-6.0/Meera/Meera.ttf
> 
>>
>> --
>> behdad
>> http://behdad.org/
> 
> 
> 

-- 
behdad
http://behdad.org/