[XeTeX] new version of HarfBuzz

[Jonathan Kew]

On 21/04/2017 19:53, maxwell wrote:
> On 2017-04-21 13:46, Jonathan Kew wrote:
>> I believe that xetex should be unaffected by that bug, because that
>> was an issue with the harfbuzz interface to graphite, but xetex uses
>> graphite directly without going through the harfbuzz layer (provided
>> you load the font with the /GR option, or whatever the fontspec
>> equivalent of that is).
>
> There is certainly a bug in the way certain sequences of characters are
> output by xe(la)tex.  Whether the bug is in xetex, or harfbuzz, or the
> font itself is beyond my ken.  Martin Hosken (who submitted the harfbuzz
> bug fix) seemed to think that would fix the issue in xetex, but I'm not
> sure he actually tested it with xetex, which I suppose would require
> re-compiling xetex.

It sounds from some off-line discussion as though it does in fact fix 
the issue in xetex, as the shaping back-end was refactored to handle 
graphite via the harfbuzz wrapper some time ago -- my initial response 
was based on how the code used to be, back when it was first written. 
Sorry for misleading!

So pulling a harfbuzz update into texlive/xetex and rebuilding would 
indeed be beneficial.

JK

>
> I can supply a simple XeLaTeX file for testing; it requires having the
> Awami Nastaliq beta font from SIL.  I can also put you (or anyone else)
> in touch with the other developers of that font, although I suppose
> Martin is the expert on this issue.
>
> I can also take this off-line if that would be more appropriate.
>
>    Mike Maxwell
>