Ok cool. I should have some time later today or tomorrow to help take a look at it too.
Jon On Mon, Dec 17, 2018 at 3:10 PM Jean-Louis Monteiro < [email protected]> wrote: > No did not ended up pushing anything. > I can continue on it. > > I was looking into some PRs and an OWB release > -- > Jean-Louis Monteiro > http://twitter.com/jlouismonteiro > http://www.tomitribe.com > > > On Mon, Dec 17, 2018 at 3:54 PM Jonathan Gallimore < > [email protected]> wrote: > > > Jean-Louis - did you have any luck? Can I help at all? > > > > Cheers > > > > Jon > > > > On Fri, Dec 14, 2018 at 4:51 PM Jonathan Gallimore < > > [email protected]> wrote: > > > > > > @Jon: instead of the logs, could we instead rely on latches? > > > > I was thinking about wrapping up the Tracker (delegate pattern) and > > use a > > > > latch so the test could actually way and avoid some timing issues. > > > > > > That sounds like a great approach. That test and associated fix came > from > > > some pretty intensive work with Websphere MQ. If the auto connection > > > tracker code changed, it would probably be worthwhile me turning that > > setup > > > on and running that stress test again. > > > > > > Jon > > > > > > > > > On Fri, Dec 14, 2018 at 4:18 PM Jean-Louis Monteiro < > > > [email protected]> wrote: > > > > > >> Starting a new thread so hopefully it makes it easier to track and > > follow. > > >> > > >> So looks like on Buildbot AutoConnectionTrackerTest is failing > randomly > > >> > > >> [ERROR] Failures: > > >> [ERROR] AutoConnectionTrackerTest.test:198 expected:<1> but was:<0> > > >> > > >> Thanks Roberto for reporting. > > >> I also ran this test many times locally and it works. > > >> > > >> I have looked into the code and I see possible issues > > >> - test is relying on logs so even with async=false, is there a way > > >> something could not be flushed out. > > >> - System.gc() maybe does not fully clean up connections > > >> - Tracker did not get time to write down the log statement > > >> > > >> Maybe others. > > >> > > >> I'm proposing here to attempt adding some log statements to see if we > > can > > >> have some pointers. > > >> > > >> @Jon: instead of the logs, could we instead rely on latches? > > >> I was thinking about wrapping up the Tracker (delegate pattern) and > use > > a > > >> latch so the test could actually way and avoid some timing issues. > > >> > > >> > > >> > > >> > > >> -- > > >> Jean-Louis Monteiro > > >> http://twitter.com/jlouismonteiro > > >> http://www.tomitribe.com > > >> > > > > > >
