Re: [ft] Glyph hinting problem with TT hinter

[Rogier van Dalen]

On 3/30/06, Werner LEMBERG <address@hidden> wrote:
> Well, this is by far the most noticeable bug in the MS TT rasterizer.
> It's a tragedy that we know that we are right and MS is wrong but the
> rest of the world thinks exactly the opposite :-(

I'm quite used to the feeling. :-)

> > At some point someone might want to adapt the FreeType interpreter
> > to mimic all MS rounding errors.
>
> I fear that too.  Any idea how to do that?

At some point I made this font "TestInstructions"
(<http://home.kabelfoon.nl/~slam/fonts/fonts.html>) that does unit
tests on rasterisers and uses oldstyle LCD-like glyphs to test some
instructions. The current version of the font contains only tests for
move instructions with points in the twilight zones, I believe. Using
these tests to fix one little bug in the FreeType rasteriser (to make
my own font work correctly :-) ) took quite a while. One could in
theory think of loads of tests for rounding. But: you'd need to
completely cover all border cases. It will take ages to do correctly
without introducing regressions. One should produce some system that
checks for regressions in rendering of all well-known fonts at all
possible sizes. Still, this scheme doesn't really test rounding in the
aliased rasteriser (as in the recent Verdana "7" case).
Personally I seem to suffer from a severe lack of time (see the "this
page was last updated on..." on my home page). Maybe these thoughts
will inspire someone though.

Regards,
Rogier