Re: [Devel] pcfread.c accelerator and size patch

[Detlef Würkner]

address@hidden (Keith Packard) wrote:

> Instead, of course we should use the pixel size of the fonts as 
> advertised in the properties (given none, we should use the above 
> computation as a fallback).

> Again, use the same size for width and 
> height, given the current semantics for FT_Set_Pixel_Sizes.

My (old 2.0.5) html docs say that for fixed-size faces
FT_Set_Pixel_Sizes() should be used instead of FT_Set_Char_Size(),
that FT_Set_Pixel_Sizes() has the parameters face, pixel_width and
pixel_height in integer pixels, and that the face object of fixed-
sized fonts contains the available_sizes array with width and height
in pixels (which IMHO implies that these are the only allowed
width/height pairs to be used).

My Amiga font engine uses this API as documented and has no problems
with the asymmetric width and height values of the winfont driver
(and my favorite version of the pcf driver), with symmetric width
and height and fixed-width fonts it would create wrong spacing
information.

<Sarcasm>
Of course I could add just another special handling mechanism based
on the name of the driver and use the face->accel values for "pcf".
There is already a "yMin = face->bbox.yMin >> 16" for "type1" and "t1cid",
a "force autohinter as default" for "type1", "t1cid" and "cff", a "max
Unicode codepoint" for each driver, special parsing of OS/2 tables,
type1/t1cid fontinfo structures, face flags, family_names and driver names
just for finding out if a "Bold" or "Regular" is part of the real font name
or not, and a "CanUseAFMfiles" flag based on driver name.
</Sarcasm, sorry..>

Why use cubic ppEM and mimic the sfnt driver? The embedded bitmaps are
part of a scalable font, and this is accessed using different metric
computations than for a bitmap font, the embedded bitmaps are just
a "goodie" you cant request explicitly. You already need special metric
computations for the winfont bitmaps which could work unchanged with pcf.
Or try to change the winfont driver to cubic ppEM...

> Sorry for the confusion.

Me too :)

Ciao, Detlef
-- 
_ // address@hidden
\X/  Detlef Wuerkner, Langgoens/Germany