On Thu, 1 Aug 2024 18:02:05 GMT, Andy Goryachev <ango...@openjdk.org> wrote:
>> I'm not sure this test needs to be that precise. And I'm fine with the >> threshold falling below 2.0/255.0 since it should be as small as possible (I >> even experimented briefly with lower numbers). I haven't run into an issue >> on the three platforms I've run this test on. > > Then the comment might not be technically correct, because it is possible to > have "two bit difference" in the integer color values, yet fail due to small > error intrinsic in floating point operations. > > Example: > > a=253 b=251 delta=0.007843137254902044 tol=0.00784313725490196 > > delta=0.007843137254902044 > tol=0.00784313725490196 (2.0 / 255.0) I've updated the constant as you suggested. ------------- PR Review Comment: https://git.openjdk.org/jfx/pull/1473#discussion_r1700739626
