On Fri, 12 Jan 2024 17:47:23 GMT, Christoph Langer <clan...@openjdk.org> wrote:
> Maybe the usual control flow would call the wrapping native method > DWMIsCompositionEnabled() from somewhere else initially such that the > initialization of Win32GraphicsEnvironment would not go through initScreens > or similar. I have not worked out how this can actually happen. I've built JDK in fast-debug mode and added java.awt.headless=true and added a bunch of printfs to be sure the code is going where I think it is but I can't make this test fail. I am running my current tests locally on Windows 11, but our CI testing does run this on Windows Server (various versions, 2016, 2019 and 2022) The only way I can make this fail is to ACTUALLY add a 2nd JNI call in there - I was beginning to wonder if somehow we were missing the WARNING, but in fact that is working. So your code path is different but what is the exact reason ? That is what we should understand. Maybe it will be 'ho hum, now I see' or maybe it will point to something we need to address. ------------- PR Comment: https://git.openjdk.org/jdk/pull/17404#issuecomment-1894497216
