Why do you think there is no consensus? I have no clue that by default failure handler should fail all test.
чт, 20 сент. 2018 г. в 21:10, Andrey Kuznetsov <stku...@gmail.com>: > I've created [1] to address this. > > Dmitriy, I like your idea of creating special test-scope handler. But there > is no consensus about it, so I don't want to rely on that potential handler > right now. We can switch to it later, of course. > > [1] https://issues.apache.org/jira/browse/IGNITE-9660 > > чт, 20 сент. 2018 г. в 20:03, Maxim Muzafarov <maxmu...@gmail.com>: > > > Andrey, > > > > I like your idea. > > > > After changing the default node failure handler to the new one we should > > carefully review the whole new test failures. For instance, calling this > > method in tests should not lead test to the node being stopped: > > > > <strong>FOR TEST ONLY!!!</strong> > > TcpDiscoverySpi#simulateNodeFailure > > > > BTW, I would like to remove this method at all from production code. > > > > On Thu, 20 Sep 2018 at 19:43 Dmitriy Pavlov <dpavlov....@gmail.com> > wrote: > > > > > But the totally ideal situation would be finding a way to fail the test > > by > > > default, not only stopping a node. > > > > > > Some time ago I've created > > > https://issues.apache.org/jira/browse/IGNITE-8227 to > > > find out a way to do so. > > > > > > чт, 20 сент. 2018 г. в 19:40, Dmitriy Pavlov <dpavlov....@gmail.com>: > > > > > > > ++1 > > > > > > > > чт, 20 сент. 2018 г. в 19:39, Andrey Kuznetsov <stku...@gmail.com>: > > > > > > > >> Igniters, > > > >> > > > >> While running tests I see a lot of ignored critical failures caused > by > > > >> {{NoOpFailureHandler}} that we use by default. In some tests, of > > cource, > > > >> critical failures are the part of normal workflow, but in the > majority > > > of > > > >> tests they indicate bugs. By using {{NoOpFailureHandler}} we just > hide > > > >> these bugs from ourselves. > > > >> > > > >> What do you think about changing default handler to > > > >> {{StopNodeFailureHandler}} at {{GridAbstractTest}} level? This could > > be > > > >> overridden in subclasses. > > > >> > > > >> -- > > > >> Best regards, > > > >> Andrey Kuznetsov. > > > >> > > > > > > > > > -- > > -- > > Maxim Muzafarov > > > > > -- > Best regards, > Andrey Kuznetsov. >
