Re: More patches for t2gload.c

[David Turner]

Hi,

> 
> Oh, I see.  The reason why you were keeping it as 16.16 during the
> parse phase is for other operators that require 16.16 operands.  You
> only shift left when assigning to a point.  Now I begin to winder if
> even that is a good idea, because some of Adobe's OpenType "Pro"
> (AJensonPro, AGaramondPro, etc...) use 16.16 numbers for points on a
> path (instead of just for side bearings and adavnced widths).
>
I suppose that these numbers come from the tool that was used to
generate the fonts. They represent, for a typical font with an EM of
1000 units, an average error of 0.5*1/1000 = 0.0005 = 0.05%, which
isn't something that can easily be spotted by the human eye, except
maybe in extreme cases..

> As for pure CFF Multiple Masters: most (all?) distiller like products
> don't embed the complete MM font; they use an instance that has been
> interpolated from the original, so we just have a plain old Type 1
> font converted to CFF.  I have never seen a PDF file with an embedded
> *full* MM font.
> 
Excellent, we can fully ditch the MM support then :-)

- David