Re: [Devel] Curious regression in ftview on MacOS X

[Werner LEMBERG]

> I saw this problem awhile back when I was adding rgb32 support to
> the mac version of the graphics surface, and found it to be due to
> the assumption in the existing mono- and gray-to-RGB32 code that RGB
> bytes are stored first, with the unused byte following (i.e. RGBx).
> Since on MacOS X, the GDI requires that for RGB32, the unused byte
> must come first (i.e. xRGB), the easiest solution was to offset the
> MacOS X GDI base address by -1; [...]
> 
> Well, now 256-gray glyphs are rendered with the GBlender code, which
> is very nice, but its code stores RGB32 as xRGB already, so my GDI
> offset is wrong for that case.
> 
> However, the original mono-to-RGB32 conversion is still used when
> antialiasing is turned off, which still assumes an RGBx alignment.
> 
> So we probably ought to resolve this; suggestions?

What about a union and macros to access its elements?  Can you provide
a patch?

> PS.: There are two bugs in grblit.c:blit_mono_to_rgb32(), lines 453
> and 459, which should be "_read++", without which the
> non-antialiased glyph doesn't render properly anyway; my apologies
> for not having noticed this when submitting the original patch.

Applied, thanks.


    Werner