On Thu, 19 Dec 2019 15:33:09 GMT, Jeanette Winzenburg <faste...@openjdk.org> wrote:
>> Florian, basically it's a single class isn't it? If so, it might be >> acceptable to add that class to the test.xx.infrastructure package (don't >> know if doing so would require a jbs issue) > > for now, you could have a look at ProgressIndicatorTest (or ProgressSkinTest) > attemptGC - might not be optimal (don't know enough about the dirty details > of gc :) but gets the test failing/passing before/after fixing a memory leak The point of having this GC-magic in a library is, that it's verified by the library that it's implementation is stable for all common JavaVersions. The attemptGC is basically bad practice. I think in the JavaFX-Codebase this kind of GC-Tests are invented about 10 times. No one knows which of these implementations are stable and their code is usually unreadable. I know it, I have done it multiple times. Copying to Infrastructure would be fine for me. It removes some benefits, like easy upgrades, Travis-verified implementation etc. But it's probably the best short-term solution. ------------- PR: https://git.openjdk.java.net/jfx/pull/71