Re: [ft-devel] A review of ttfauthint by designer of Siri

[vernon adams]

On 10/02/12 09:24, Werner LEMBERG wrote:
> > > It's rather that code is missing which checks that segments from
> > > different outlines have enough vertical separation.  I have to
> > > investigate yet whether this is a problem in my bytecode or a
> > > general problem with the FreeType autohinter.
> >
> > It's allways been noticeable to me that when i manually hint fonts
> > on Windows GDI, the font at small pt sizes renders on Freetype
> > sytems with the x-height and Caps height lower than under Windows
> > GDI.
>
> A *hinted* font?  This sounds strange.  Bytecode executed with
> FreeType and the MS rasterizer should give the same result.  If you
> have evidence for this behaviour please give more details since it
> sounds like a bug.

Werner,
Not sure if we are talking about the same thing, and i may not know what 
i am talking about :)

Look at the 4 screenshots at http://code.newtypography.co.uk/?p=2265
click the shots to see full res images.

2 versions of same Truetype font (1 manually instructed with MS Visual 
TrueType, 1 with no TT instructions)
Both rendered under freetype 2.4.8 on KDE 4.8 with 'Full hint" and 
subpixel rendering. Top 2 shots are 9 pt font size, lower shots are 8 pt 
font.

At 8 pts the x-height of the hinted font is same as the non-hinted font.
But, at 9 pt the x-height of the non-hinted font is rendered taller than 
the x-height of the TTinstructed font. It's 1 pixel taller in fact.

My point is that the freetype autohinter should (if possible) render 
x-heights & cap heights the same as the freetype bytecode interpreter. 
Though from your comment below, i think you are saying it's not a good 
idea to force freetype to do that.


-v



> > In my opinion ttfautohint needs to use freetype libs that snap
> > x-height and caps-heights lower than the standard freetype
> > autohinter.  Does that make sense?
>
> I don't think so.  You simply can't compare manually hinted TT fonts
> with the result of the autohinter.  In general, rounding down at small
> PPEM values causes more troubles than rounding up.
>
>
>      Werner