Nikolaus Waxweiler
2018-11-10 16:56:01 UTC
Hi!
Motivated by https://bugzilla.redhat.com/show_bug.cgi?id=1645763, I made
the attached patch to use FreeType's default LCD filter as the default
one instead of the "legacy intra-pixel" filter.
This gives me some test suite failures that I'm not sure how to fix.
Should the reference images for test-antialias-subpixel.c be simply
swapped? One problem might be that FreeType as of recently has two ways
to subpixel-render:
1) The traditional one behind the FT_CONFIG_OPTION_SUBPIXEL_RENDERING
define that comes with the ability to set LCD filters.
2) A new one called "Harmony" that gets used if the define above is
undefined and gives the same result as the light LCD filter ± rounding
errors. You can't set filters here but can specify the subpixel geometry
(e.g. something for pentile displays).
So a test should probably test for one or the other and xfail otherwise?
Motivated by https://bugzilla.redhat.com/show_bug.cgi?id=1645763, I made
the attached patch to use FreeType's default LCD filter as the default
one instead of the "legacy intra-pixel" filter.
This gives me some test suite failures that I'm not sure how to fix.
Should the reference images for test-antialias-subpixel.c be simply
swapped? One problem might be that FreeType as of recently has two ways
to subpixel-render:
1) The traditional one behind the FT_CONFIG_OPTION_SUBPIXEL_RENDERING
define that comes with the ability to set LCD filters.
2) A new one called "Harmony" that gets used if the define above is
undefined and gives the same result as the light LCD filter ± rounding
errors. You can't set filters here but can specify the subpixel geometry
(e.g. something for pentile displays).
So a test should probably test for one or the other and xfail otherwise?