On Mon, 5 Dec 2022 05:29:53 GMT, Prasanta Sadhukhan <psadhuk...@openjdk.org> wrote:
> The test is problemlisted for windows so not sure if the observation "passes > on Windows" is correct.. When it failed on Linux on pre-submit task i ran it locally and deproblemlisted it on Windows. I haven't pushed the problemlist change but i found that on Windows it is unstable because of the way it was taking screenshots and it was passing because it compared the two screenshots that are not supposed to be the same and it passes because they are not but that is the false positive because the difference is not in the relevant part of the screenshot but in the border overlapped by the popup. Except this false positive is not a false positive because on Windows these components should be the same for the logic tested is only valid on Linux - selection of the text in the focused window does not clear selection in non-focused one on Windows or Mac. So in the similar fashion test "accidentally" passes on Mac but for different reason - the screenshots were different because on Mac the selection color in non-focused window is different so screenshot does not match while functionally select ion is not cleared so it is another "false but correctly positive". ------------- PR: https://git.openjdk.org/jdk/pull/11408
