On Thu, 30 Sep 2021 04:43:19 GMT, Sergey Bylokhov <[email protected]> wrote:
> The fix for 8274296 added uiScale=1 to some of the tests to fix the problem > of incorrect pixel color capture by the robot. > In both cases uiScale=1 or uiScale=2 the same call should be made to the > macOS. But if uiScale=1 is set we just read one pixel on the screen, but if > uiScale=2 is set we prepare the data in some clumsy way(at the end we should > also read just one pixel). > > Since uiScale=1 affects the test case execution means that we probably have > some bugs in the coordinate/size calculation of the area. Note that none of > these tests have dependencies on the precise rendering, so it is quite a > common use-case and may affect any external applications. > > This is a request to update the https://github.com/openjdk/jdk/pull/5687/files > and move all tests to the problem list. Once again, hi-dpi does not matter for these tests. It is not what they are trying to test and there is no other "hi-dpi" testing going on anyway on these other platforms, and I keep saying bug in macos not JDK for the reasons I wrote in https://bugs.openjdk.java.net/browse/JDK-8274106 which is tracking the underlying issue and is aware of ALL the tests that fail. I looked at the Robot code and it was getting different answers *from a macos api* depending on whether it asked for some larger region first. ------------- PR: https://git.openjdk.java.net/jdk/pull/5766
