[Gnash-dev] Re[2]: Text quality (was: Questions about trapezoids, triangle stripper etc.)

[Udo Giacomozzi]

Hello strk,

Monday, October 2, 2006, 4:38:41 PM, you wrote:
s> mm... then I guess we'd be working on it togheter.
s> I really want this to be fixed before 0.7.2.

I try to start with the rewrite of the renderer design. It's not that
easy to keep backwards compatibility for other backends.

s> There's some crappy code in fontlib.cpp. It relies on the full font
s> set being already defined, does a lot of things that I don't really
s> understand and after the fact each texture_glyph is equipped with
s> it's bitmap_info. As a first step I would dedicate on cleanup
s> of that file, so it's easier to move functionalities around.

Somewhere in fontlib it declares a shape_defunition for the glyph. It
would be important to follow that way (don't have the renderer deal
with fonts at all).

BTW, fonts are the reason why shape_definition::display() is split in
two. When a glyph is rendered, there is no character instance.


>> - Are glyph outlines *always* available, i.e. also when the font is
>>   *not* included in the SWF file, but is a system font?

s> I don't think so, device fonts are actually not supported.

When only embedded fonts are supported, then glyph outlines should
always be available. Probably Gnash does not even require FreeType,
then?

As for the AGG backend, the rendering of small glyphs is not that bad
I think. See the "glyph-test" at http://www.indunet.it/temp/gnash/

It matches the quality of the "Anti-alias for animation" mode of
flash. No idea how a "Anti-alias for readability" could be implemented
(difference is visible at small text sizes).

For other renderers we still have to create bitmaps and I don't know
how that's done at the moment.


Ok, I try to write a first version of a redesigned backend so that it
may be committed to CVS. The tricky part is the backwards
compatibility.

Udo