On Wed, Aug 30, 2017 at 2:13 AM, Enrico Olivelli <eolive...@gmail.com> wrote:
> 2017-08-30 11:05 GMT+02:00 Sijie Guo <guosi...@gmail.com>: > > > On Wed, Aug 30, 2017 at 1:42 AM, Enrico Olivelli <eolive...@gmail.com> > > wrote: > > > > > Thank you Sijie I forgot to subscribe to new issues@ mailing lists > and I > > > missed some PR/Issue > > > It is a very good enhancement > > > > > > when we are stable I would like to switch the code-coverage tools to > this > > > new environment. > > > > > > > I am not sure why do you need to switch. It is just running exact same > > maven command using an official maven docker image. > > > > Also, code-coverage should be part of nightly release ci. I don't see a > > reason to have a separate code-coverage ci. > > > > OK, I will do that once the code coverage job is stable > > > > > > > > > > > > > I think that on Travis-CI we are already executing in containers, but > > > Travis does not give us enough resources to run the complete suite > > > > > > > It is because currently we are use 'reuseFolk=false', the whole suite > need > > to run very long time because it has to folk a jvm for individual test > > cases. > > > > We need to have better clean ups for tests, I hope we will enhance the > suite soon. > It would be great to run parallel tests too, now that we have ephemeral > ports it would be feasible > It is not the port problem. It is a problem from open fds when reusing jvm forks. > > > > > > > > > it is a pity because I would like to integrate the code-coverage on PRs > > > > > > How do you want to use code-coverage for PRs? > > > > I see in some open source projects, like Pravega (which does not use > Coveralls), that there is bot which runs tests and code coverage tools and > adds comments to the PR which tell something like 'this patch will > increase/reduce code coverage by XX %' > this will be a good enhancement > Okay. but what does it actually mean to the contributors? What do the contributors need to do if code coverage is reduced by XX%? - Sijie > > -- Enrico > > > > > > > > > but > > > I think we will need Travis to make things simple, otherwise I will > > figure > > > out how to obtain it with jenkins > > > > > > > > > > > > > > > > > > > -- Enrico > > > > > > 2017-08-30 10:34 GMT+02:00 Sijie Guo <guosi...@gmail.com>: > > > > > > > Hi Enrico, > > > > > > > > You can check the description in the pull request - > > > > https://github.com/apache/bookkeeper/pull/481 > > > > > > > > We've tried to fix the flaky tests as we can. However there are some > > test > > > > cases that I can't figure out what are the root causes and seems to > be > > > > related to network settings. > > > > #481 is adding a script to run maven build in a docker environment > and > > we > > > > setup another pull request CI job, so a pull request is running on > > > jenkins > > > > machine and also in docker. > > > > We can monitor and compare to see if what is really happening. > > > > > > > > If running in docker can produce a much stable CI environment, we can > > > > remove the one that runs on physical CI machine. > > > > > > > > - Sijie > > > > > > > > On Wed, Aug 30, 2017 at 12:17 AM, Enrico Olivelli < > eolive...@gmail.com > > > > > > > wrote: > > > > > > > > > Hi, > > > > > I have just seen this new job bookkeeper-precommit- > > pullrequest-docker > > > > > and the commit about tests on docker + the change about running > > > zookeeper > > > > > with real ephemeral port. > > > > > I think these are great changes ! > > > > > I have missed that piece of the story, most of this work is related > > to > > > > > https://github.com/apache/bookkeeper/issues/463 > > > > > > > > > > @Sijie @Jia > > > > > can you please summarize the changes and the direction ? > > > > > > > > > > Cheers > > > > > Enrico > > > > > > > > > > > > > > >
