On Fri, 24 Sep 2021 18:09:38 GMT, Sergey Bylokhov <[email protected]> wrote:
>> As described in https://bugs.openjdk.java.net/browse/JDK-8274106 a number of >> tests fail due to some inaccuracy in pixels copied from the screen when >> scaling down from device space to user space. >> The subject of most of these tests is such that uiScale isn't important and >> we can have them pass with -Dsun.java2d.uiScale=1. >> For the rest we can problem list them - as narrowly as we can on >> macos-aarch64 which is the only place we've seen the problem (not happening >> on retina displays with x64 it seems) whilst we try to figure out the root >> cause which is either in Apple code or something we are doing to trigger >> that Apple API misbehaviour. > > test/jdk/java/awt/Dialog/SiblingChildOrder/SiblingChildOrderTest.java line 29: > >> 27: * @summary [macosx] Order of overlapping of modal dialogs is wrong >> 28: * @key headful >> 29: * @run main/othervm -Dsun.java2d.uiScale=1 SiblingChildOrderTest > > Switching this tests to scale=1 seems completely wrong. On macos it just > tweak the rendering so that bug became invisible due to lost of precision. > And now we will run the tests in lowDPI on all other platforms. If really needed I can file a follow-on to change this test to problem listed on macOS like some of the others .. but this test doesn't look to me like its a pixel precision test ... ------------- PR: https://git.openjdk.java.net/jdk/pull/5687
