Prior to this change, the optimization for filling 1x1 pixmaps that simply replaced the destination pixmap with the foreground color pixel was applied regardless of the bitwise logic operations, leading to incorrect results. This caused the libreoffice_xrgb test in rendercheck to fail on systems using EXA with a non-24bpp screen depth, such as is possible with the mach64 driver.
This change amkes it so the optimization is only applied if the bitwise logic operator is a direct copy, which enabled mach64 with 24bpp to pass the libreoffice_xrgb test in rendercheck.
I apologize if I have made any mistakes in regards to style, function, or the submission of this merge request. If there is anything I can do to improve in the future, please let me know.
Thank you for looking over this and for any reply.
Signed-off-by: George Matsumura firstname.lastname@example.org @daenzer @anholt