Re: [ft] Subpixel Hinting Patch

[James Cloos]

>>>>> "WL" == Werner LEMBERG <address@hidden> writes:

WL> However, it seems that you (completely?) replace the current
WL> gray-level hinting stuff with cleartype-like hinting.  Is this
WL> intended?  Shouldn't this be three different `modes'?  Bi-level,
WL> standard gray-level, and cleartype-like?

[ I was going to reply directly with this, but Werner's question is
  too relevant not to quote. -JimC ]

I tried out the patch, installing it as my system freetype.

My fontconfig setup suggests the use of the legacy filter for instructed
glyf fonts, and the default filter for autofit and light-autofit fonts.
I have cairo patched to follow that suggestion, so gtk apps, in particular,
do just that.  I beleive qt apps also do so, but i may be mis-remembering
that.  XFT, on the other hand, continues to do its own thing, which is what
the legacy filter was designed to emulate.

And my primary UI fonts are DejaVu Serif and DejaVu Sans Mono, sized such
that the stems are one pixel wide when instructed.

With you patch, the fonts look very poor.  It increases colour fringing
to the point of distraction.

I even tried with the default filter instead of the legacy filter.

I very much like the idea, but I had to go back.

I think the new hinting should be in addition to, and not instead of
the current code.

Part of the issue, of course, is that the fonts I care about and the
fonts you are testing against are almost disjoint sets.

So before going back I did run ftview an ftdiff on a few of the fonts
which came with the laptop.  (It arrived with XP.)

The most interesting result of that test was that PALA.ttf (Palatino
Linotype, version 1.40) was invisible with your patch.

-JimC
-- 
James Cloos <address@hidden>         OpenPGP: 1024D/ED7DAEA6