Re: Courier-Oblique

[David Turner]

Hi Tom,

> 
> I took a look at how FT_Load_Glyph is implemented, and saw that it uses the
> autohint module (if one is available for for the current face's driver).
> 
> So, I was mistaken several times over this past weekend.  The problems weren't
> in T1_Load_Glyph, nor with metrics handling.  The problems were elsewhere, as
> you have noted.
> 
Actually, it's a combination of several factors. basically, the Type 1
"font transform" is not the same thing as the "face transform" that is set
through FT_Set_Transform.

the problem comes from the fact that in the current Type 1 glyph loader,
the font transform is not applied when FT_LOAD_NO_RECURSE is called.

We could activate it, and we would have auto-hinted transformed fonts
instantly. However, this doesn't bode well in the point of view of consistency,
given that one would normally expect that an auto-hinted obliqued font looks
like its corresponding obliqued auto-hinted font.

the only way to do that properly is performing the auto-hinting before
the font transform, then applying the later. we need to modify the glyph
loader to return the font transform to the auto-hinter then..

of course, the "face transform" can be applied after all this anyway..

> Out of curiousity, wouldn't it be better to handle the hint calculations 
> *after*
> transform?  The reason why I ask is that in the case of obliquing, the 
> vertical
> hints might change position (though not width, etc...).  The horizontals hints
> won't suffer from this, though.  Or are hints "attached" to points on the 
> curve,
> so that even if the points change position, the hints remain valid?
> 
hints are computed from the points coordinates in the auto-hinter, and the 
directions
of segments. applying a transform _before_ auto-hinting can result in certain 
cases
in greatly different vertical hints, and thus a loss of consistency between the
regular and obliqued version of the font.

Regards,

- David