Without looking at specifics, I think it would be awesome to have real tests for the GUI. Cayenne is dependency-free where it matters - when it is embedded in the apps. Using a third-party lib to test the code during the build does not affect that in any way. So we should be fine.
Andrus > On Dec 5, 2017, at 5:54 PM, eme...@gmail.com wrote: > > Hi Dev@Cayenne community. > > During recent months after an intensive work with the modeler, I found a set > of bugs related to visualization, two of them have been already fixed > (CAY-2379 [1] & CAY-2384 [2]) , currently I'm in the process for reporting > and solving at least 3 new ones. > > During the fixing I have implemented some GUI testing specific for the > modeler, using assertj-swing [3]. This has passed all validations for the > building process, including Travis-CI by using xvfb [4] for all the active > profiles in master branch. Also it requires including the assertj-swing[5] > dependency in the parent-pom with test scope. > > So, I wonder if including this new GUI testing feature would be aligned to > apache cayenne goals; I remember reading recently from Andrus his happiness > posting about "Cayenne Core is Dependency-Free" [6] as part of the version > 4.X key improvements. > > Although the implemented GUI testing does not add too much time to the > building process, all validations until now look stable, and by include this > kind of testing we will be working toward guaranty a better end-user > experience, I would like to ask for your opinions before keep working on that. > > Regards > > EmeCas > > > [1] https://issues.apache.org/jira/browse/CAY-2379 > [2] https://issues.apache.org/jira/browse/CAY-2384 > [3] https://joel-costigliola.github.io/assertj/assertj-swing.html > [4] https://www.x.org/releases/X11R7.6/doc/man/man1/Xvfb.1.xhtml > [5] http://mvnrepository.com/artifact/org.assertj/assertj-swing/3.8.0 > [6] > https://medium.com/@andrus_a/apache-cayenne-4-1-java-8-no-dependencies-smaller-memory-footprint-629eb47a725f > >