Re: [ft-devel] Can freetype be more tolerant with this font?

[Alexei Podtelezhnikov]

> > > I obviously have no idea how important the check for FT_CURVE_TAG_CUBIC 
> > > is, just that for this particular font removing it makes it work.
> >
> > Below is the dump of the m outline data. The contour 1 contains a run
> > of four control points FT_CURVE_TAG_CUBIC (0x2), which is not how a
> > Bezier curve would be specified. I think Freetype is correct to reject
> > this font.
>
> Ok, so the font is not correctly formed, but it seems that freetype can still 
> render it "just fine".

I don't disagree but I hesitate to create expectations. The second
order (truetype) curves have well established conventions that any
sequence of on- and off-points is unambiguously interptretable. Hence
FreeType would never fail to render truetype. The third order curves
(type1 etc) have many arbitrary restrictions: the first point has to
be on and no implicit points allowed. This restrictions seems
unnecessary, but there must be a discussion and an agreement in what
to do with a run of control points or when there is just one control
point given. Keep in mind that FreeType permits mixing the second and
the third order curves. IT gets too complicated. Do we really need to
bend over backwards?

By the way, does Adobe render this file properly? That would be a
strong argument to support it too.


> > Contour   1
> >  92   x:   456   y:   178   t: 0x1
> >  93   x:   484   y:   358   t: 0x1
> >  94   x:   484   y:   358   t: 0x2
> >  95   x:   484   y:   358   t: 0x2
> >  96   x:   484   y:   360   t: 0x2
> >  97   x:   484   y:   360   t: 0x2