On Mon, 8 Jul 2024 22:40:42 GMT, Kevin Rushforth <k...@openjdk.org> wrote:
>> even with max=200 we might see occasional failures due to load or other >> circumstances. >> >> I wonder a better approach would be to check whether the measurement of the >> delay falls sufficiently outside of the default 1000 ms (which in itself >> might be more than that)? > > We have many tests that will occasionally fail with a heavily loaded system. > > The question is whether we can come up with a max delta that we can reliably > use that is less than the difference between the default value and the test > value. Otherwise, we can't distinguish them. > > On two different test systems, both of which fail pretty consistently with > 50 msec, I see a 100% pass rate over several tries with 150 msec. More > testing is needed. Another possibility is the code that measures the time it takes to show the tooltip. The current code uses Util.waitForLatch(), L244 which was written for a different use case and actually introduces a small measurement error as it includes the time needed for the context switch. Perhaps a better solution would be to note the timestamp in L271 where the tooltip listener gets called (similar to the way the start moment is captured in L241) ------------- PR Review Comment: https://git.openjdk.org/jfx/pull/1394#discussion_r1669418098
