https://bugs.documentfoundation.org/show_bug.cgi?id=160036

--- Comment #12 from Patrick Luby (volunteer) <[email protected]> ---
(In reply to Noel Grandin from comment #11)
> It would, but upgradning skia is a painful process. We replace their build
> scripts (because they use some other build tool that we don't support), and
> their API sometimes changes or move around, which means we then need to
> adapt our code.

I wonder if something changed or moved around in m116 (and likely later) that
is short-circuiting our invert and XOR strings. After all, the invert and XOR
strings work in the GPU Skia implementations, but not in the Raster
implementation.

When I get some time, I am thinking of stepping into the Skia code that handles
our XOR string in a debugger and see if I can find anything in Skia that is
failing when using Raster.

If I don't find anything with that, my fallback plan is to rewrite the XOR code
in VclPixelProcessor2D::processInvertPrimitive2D() when using Skia/Raster to
extract a BitmapEx from the OutputDevice, iterate through its pixels and XOR
each pixel, and then draw the XOR'd BitmapEx onto the OutputDevice with clip
set to the polypolygon selection shape.

Sound reasonable?

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to