Bug 34886

Created attachment 43989 [details]
file-roller, on the left official tarball; on the right with the commit reverted

Created attachment 43990 [details]
Some text in firefox. The one at the top is the official tarball; the one at the bottom is with the commit reverted

(In reply to comment #0)
> Created an attachment (id=43988) [details]
> ~/.fonts.conf
> 
> I've been experiencing some jagged-looking fonts in GTK apps with Cairo >
> 1.8.8; I've narrowed it down to changes between 1.9.8 and 1.9.10. Reverting
> this commit:
> http://cgit.freedesktop.org/cairo/commit/?id=7a023a62f7517ad0d54f4d59c99909fadcc05e82
> 
> made fonts looks better. See the attached screenshots.
> 
> I am using cairo-1.10.2, freetype2-2.4.4 (default upstream, i.e. the bytecode
> interpreter is enabled, but sub-pixel rendering is disabled (due to patents)).

That commit replaces a cairo-internal implementation of subpixel filtering with
the appropriate api calls to freetype. This respects your system settings and does
not work around your freetype library, hence to get subpixel filtering you should
use a freetype library with subpixel enabled.

Can you confirm if enabling subpexel in freetype fixes your problem?

(In reply to comment #3)
> That commit replaces a cairo-internal implementation of subpixel filtering with
> the appropriate api calls to freetype. This respects your system settings and
> does
> not work around your freetype library, hence to get subpixel filtering you
> should
> use a freetype library with subpixel enabled.
> 
> Can you confirm if enabling subpexel in freetype fixes your problem?

Yes, using freetype with enabling subpixel rendering at build time, I get similar looking fonts (similar to the "good" fonts in the attached ^ screenshots) with lcdlegacy filter.

So this is not a bug/regression? .. so now I have to enable a patent-covered feature in freetype2 to use freetype's legacy subpixel rendering, right?

Re-reading you comment again, I see that this bug is actually invalid.

Thanks :)