Re: [ft-devel] maxp.maxStackElements (e: Freetype-devel Digest, Vol 141, Issue 4)

[Hin-Tak Leung]

Hi Cosimo,

It is 32 extra elements - elements are minimum 16-bit wide? FreeType probably 
use 32-bit int's for it.

I checked my test results of the ~5000 fonts I tested for E6045 ( the "open 
font" part is up at
http://htl10.users.sourceforge.net/tmp/FontVal-test-results-2016July/ ) - the 
FontVal code for stack overflow.

You can do something like
      bzip2 -dc fedora-1.0-texlive.txt.bz2| grep E6045
to extract that entry yourself

Because of the +32,  E6045 is not seen with FontVal 2.0 at all. (I'll fix it 
one day...)
But a few fonts can trigger that with FontVal 1.0. They are

/usr/share/texlive/texmf-dist/fonts/truetype/public/fontmfizz/font-mfizz.ttf 
from texlive-fontmfizz-svn35892.0

DecoTypeNaskh.ttf and Apple Color Emoji.ttf from apple (various versions)

and a few *-*-hkscs-u.ttf CJK fonts from my private stash.

You can probably play with either the two apple fonts or the mfizz font from 
texlive a bit, move the value up and down a bit to see what FontVal 1.0 thinks 
is the correct answer.

BTW, all of the flagged errors from FontVal 1.0 happens in the fpgm! So you 
need to make sure that you are processing fpgm correctly.

I'd be interested if you can say, modify any of those fonts to get FontVal 1.0, 
FontVal 2.0 (taking the +32 into account) and VTT into some sort of agreement.

Hin-Tak 

--------------------------------------------
On Sun, 9/10/16, Cosimo Lupo <address@hidden> wrote:

 Hi
 Hin-Tak,
 Thank you for the
 insights about FontVal 2.0 and the 32 extra bytes that
 FreeTypes allocates for the stack.I've
 contacted Mike Duggan from Microsoft, maybe they could
 answer my questions.I'll keep you
 posted.Cheers.
 
 Cosimo