[HarfBuzz] Indic reordering failing with latest HB trunk

[Behdad Esfahbod]

On 13-10-24 07:58 AM, Shriramana Sharma wrote:
> On Thu, Oct 24, 2013 at 2:03 AM, Jonathan Kew <jfkthame at googlemail.com> wrote:
>>
>> Seems to work OK here.  I suspect something about your build must be broken;
>> are you getting the "fallback" shaper instead of "ot"?
> 
> I am not sure where to look for this.
> 
>> Or did you build
>> without any valid Unicode callbacks?
> 
> Unicode callbacks (you want at least one):
>         Glib:                   true
> 
> So I guess this is not the problem.
> 
> All I did was to do git pull && ./autogen.sh --with-graphite2=auto
> --prefix=/usr/local && make && sudo make install.
> 
> Anyhow the latest commit 6e613f3 seems to fix whatever it is (though I
> don't understand how the shift count fix could affect this).

Interesting.  If a shift count of 41 was being interpretted as 9 on a 32bit
architecture, I can imagine things going really bad.

> If I do git checkout ac8cd51 and recompile, I still get the bug. By
> doing git bisect I identified commit c16012e as the culprit but I have
> no idea what that commit introduced which 6e613f3 corrected. Please
> look into this a bit and identify it lest it resurface later on.
> 
> Thanks.
> 
> BTW about hb-indic -- do we still use that list?

No, we decided to retire that quite a while ago.

-- 
behdad
http://behdad.org/