Are the tests so brittle that, even with flaky, people will run upon false failures as part of contributing a PR? If so, do we have a list of the brittle ones (and the things that would disambiguate a true failure from a false failure) that we can add to the documentation?
On Tue, Oct 2, 2018 at 11:58 AM Shane Ardell <[email protected]> wrote: > I also would like to eventually have these tests automated. There are a > couple hurdles to setting up our e2e tests to run with our build. I think > the biggest hurdle is setting up a dedicated server with data for the e2e > tests to use. I would assume this requires funding, engineering support, > obfuscated data, etc. I also think we should migrate our e2e tests to > Cypress first because Protractor lacks debugging tools that would make our > life much easier if, for example, we had a failure in our CI build but > could not reproduce locally. In addition, our current Protractor tests are > brittle and extremely slow. > > All that said, it seems we agree that we could add another PR checklist > item in the meantime. Clarifying those e2e test instructions should be part > of that task. > > On Mon, Oct 1, 2018 at 2:36 PM Casey Stella <[email protected]> wrote: > > > I'd also like to make sure that clear instructions are provided (or > linked > > to) about how to run them. Also, we need to make sure the instructions > are > > rock-solid for running them. > > Looking at > > > > > https://github.com/apache/metron/tree/master/metron-interface/metron-alerts#e2e-tests > > , > > would someone who doesn't have much or any knowledge of the UI be able to > > run that without assistance? > > > > For instance, we use full-dev, do we need to stop data from being played > > into full-dev for the tests to work? > > > > Casey > > > > On Mon, Oct 1, 2018 at 8:29 AM Casey Stella <[email protected]> wrote: > > > > > I'm not super keen on expanding the steps to contribute, especially in > an > > > avenue that should be automated. > > > That being said, I think that until we get to the point of automating > the > > > e2e tests, it's sensible to add them to the checklist. > > > So, I would support it, but I would also urge us to move forward the > > > efforts of running these tests as part of the CI build. > > > > > > What is the current gap there? > > > > > > Casey > > > > > > On Mon, Oct 1, 2018 at 7:41 AM Shane Ardell <[email protected]> > > > wrote: > > > > > >> Hello everyone, > > >> > > >> In another discussion thread from July, I briefly mentioned the idea > of > > >> adding a step to the pull request checklist asking contributors to run > > the > > >> UI end-to-end tests. Since we aren't running e2e tests as part of the > CI > > >> build, it's easy for contributors to unintentionally break these > tests. > > >> Reminding contributors to run these tests will hopefully help catch > > >> situations like this before opening a pull request. > > >> > > >> Does this make sense to everyone? > > >> > > >> Regards, > > >> Shane > > >> > > > > > >
