Re: [ft-devel] Re: [Devel] Kerning problem with capitals

[Werner LEMBERG]

> Playing with Fontforge I saw a message that default EM-size for
> postscript is 1000 where this font has 2048. This font saved as new
> TTF with EM 1000 gives proper results in pdflatex[3] - fragments
> labeled with F1/F3 are written with gotykposzarpany1[2](old w/o
> kerning) and with problematic font gotykposzarpany3[1].
>
> I tried this bad font[1] with ftstring (FreeType 2.1.10) with the
> same problem - see left-bottom window on screenshot[4]. O this image
> upper row is with font transformed to EM1000, lower original with
> EM2048. Left side - ftstring, right fontforge.

Gotyk3.ttf is really broken IMHO!  And I think that FontForge's
behaviour is not fully correct also (George, please comment).  First
of all, ftstring handles kerning correctly; the values it returns are
properly scaled.

The very problem are the horizontal metrics of the glyphs.  Let's have
a look at glyph `d'.  The `glyf' table gives these dimensions:

    xMin = 188
    yMin = -2
    xMax = 963
    yMax = 1388

and the `hmtx' table yields

    width = 812
    lsb   = 28

According to the TT specs, the right side bearing can be computed as

    rsb = width - (lsb + xMax - xMin)

For glyph `d' we thus have rsb=9, but FontForge displays rsb=-151
(which is apparently `width - xMax').

If you change that manually in FontForge for all glyphs, you get
exactly the same result as with ftstring.

I really wonder whether this font displays correctly on MS Windows, or
whether my conclusions are wrong...


    Werner