Re: [Gnash-dev] device fonts

[Udo Giacomozzi]

Hello strk,

Tuesday, June 12, 2007, 6:54:56 PM, you wrote:
s> We should decide if it's worth fixing the bugs with OGL and core
s> *and* to implement draw_bitmap for AGG or focus on creating
s> vectorial glyphs instead.

I don't think implementing draw_bitmap for AGG would be a problem.
However, how do these fonts work? Is there *one* bitmap for each
letter which has to be scaled, rotated, downsampled? In that case it
would result in very bad rendering quality of fonts (and after all,
fonts need to be rendered well to be readable at small sizes).

This comes close to the discussion I'm having in the AGG mailing list,
which is about implementing bitmap caches for vector shapes in AGG. It
seems that it is not feasible except for caches that are valid for one
exact position, scale and rotation (ie. for non-moving objects).


s> Vectorial glyphs are basically
s> shape_character_def so in order to support them we should convert
s> the freetype glyph to a shape character. Drawing primitive are
s> about ready (there for Drawing API) so we'd just need to find
s> out how to perform conversion (libfreetype seems to support it).

Don't think we need the drawing API for that (instead create paths
directly) and it should not be difficult. I don't remember where, but
I saw a document describing exactly this (converting from TTF to SWF).
Basically all that needs to be done is converting the cubic bezier
curves used in TTF ("type1") to quadratic bezier curves used in Flash,
which means adding a anchor point exactly in the middle of the two
cubic curve anchor points.

Sorry, I haven't the URL at hand now..




Udo