On Tue, 27 Oct 2020 20:40:23 GMT, Kevin Rushforth <k...@openjdk.org> wrote:
>> I've added runFinalization followed with another System.gc (otherwise it >> wouldn't make any difference) >> I'm quite sure it's avoiding an error in the library, but I don't have a >> unit-test for it yet. >> >> It's a tradeoff between speed and reliability. The library tries to avoid >> false-negative test results at all costs. >> false-positive results are tolerated for speed if they are "sufficiently >> reliable". >> The Library has tests for assertNotCollectable, which are executed with >> every commit for different JVM's, and I don't remember any false-positive >> results (which would be visible as a negative test result, because the >> unit-test checks for failing tests). For that reason, we can assume they are >> "sufficiently reliable" despite the theoretical problem. > > That sounds fine then. > > As soon as you've pushed the follow-on fix (to add the `runFinalization` and > a second `gc`) I'll finish my review. I forgot to push my changes yesterday. They are pushed now! ------------- PR: https://git.openjdk.java.net/jfx/pull/204
