[jfx-tests] RFR: 8315842: 3D tests fail because of edge pixel differences

[Nir Lisker]

On Thu, 7 Sep 2023 10:17:28 GMT, Jayathirth D V <jdv at openjdk.org> wrote:

> Out of 62 3D tests, 26 tests fail because of minute color differences in edge pixels.
> These tests are used to verify 3D rendering with different parameters like translation, rotation.
> 
> So adding little color tolerance will not change the test behavior and allows us to use these tests to automatically verify any regression introduced in 3D rendering.
> 
> Added 5% color tolerance and with this change 23 of these tests pass.
> 
> Some sub-tests under below 3 tests continue to fail because of other reasons:
> [test/scenegraph/fx3d/camera/fixedeye/PerspectiveCameraFixedEyeIsolateTest.java](file:///Users/jdv/dev/workspace/jfx/jfx-tests/functional/3DTests/build/test.workdir/test/scenegraph/fx3d/camera/fixedeye/PerspectiveCameraFixedEyeIsolateTest.jtr)
> [test/scenegraph/fx3d/camera/parallel/ParallelCameraIsolateTest.java](file:///Users/jdv/dev/workspace/jfx/jfx-tests/functional/3DTests/build/test.workdir/test/scenegraph/fx3d/camera/parallel/ParallelCameraIsolateTest.jtr)
> [test/scenegraph/fx3d/camera/perspective/PerspectiveCameraIsolateTest.java](file:///Users/jdv/dev/workspace/jfx/jfx-tests/functional/3DTests/build/test.workdir/test/scenegraph/fx3d/camera/perspective/PerspectiveCameraIsolateTest.jtr)
> 
> Also i see that some of the camera tests just draw white images, this also needs to be verified.
> With this change 41 out of 62 3D tests will run properly.

Not sure if it's relevant, but when I wrote a 3D test with a color tolerance value it turned out to be too low on Mac (but not on Windows), as Kevin pointed out to me: https://github.com/openjdk/jfx/pull/43#discussion_r502049390. We ended up using his suggested value (10/255 =~ 0.04) because it was good enough, but it could probably be a bit lower.

-------------

PR Comment: https://git.openjdk.org/jfx-tests/pull/5#issuecomment-1711859852