@Marco: Hehe totally agree with the approach. I just don't want to make those reports looks better, I want to be sure that the code that we have is good.
Oh yeah: coverage reports.
IMHO, coverage reports are like weather reports: good to check now and then, but not something you want to plan around.
Also, wrt agile/efficiency, we're better off making sure that the tests do actually test what we want them to -- which is pretty much exactly what you've been doing, Mauricio.
11-05-12 11:36, Maciej Swiderski:+1 for cleaning up the test, it takes quite some time to run all of them.
Regarding changes of package good to keep in mind:
- installer stuff that it declares DefaultUserGroupCallback in both web.xml and jbpm.usergroup.callback.properties.
- human task war tests could declare...
There should be LDAPUserInfo class that could go into identity package as well.
Coverage report look quite ok :) and improvement is always welcome.
Cheers
Maciej
2012/5/11 Mauricio Salatino <salaboy@gmail.com>
Sure, I will do that if all the other points make sense, and I will not push the commented out tests :)
On Fri, May 11, 2012 at 12:04 PM, Marco Rietveld <mrietvel@redhat.com> wrote:
Hi Mauricio,
Feel free to delete the code instead of commenting it out, especially if you can't use @Ignore.
Commented out code is more often confusing than not. :)
Thanks,
Marco
11-05-12 10:46, Mauricio Salatino:Hi guys,I was working yesterday trying to understand why we have 390 test running inside jbpm-human-task-core and if they are really necessary.This is my initial approach so forgive me if I'm missing something, the following notes and changes are what I found until now:
1) All the test classes using directly TaskClient should be removed, now everything should go through TaskService and AsyncTaskServicetest which are placed inside async and sync. For that reason all the test inside org.jbpm.task.service are commented out.I've checked that all the test classes that were using TaskClient now has their correspondency inside the org.jbpm.task.service.base.async package.Where the only important change is from TaskClient to AsyncTaskService. We were running the same tests for both wasting a lot of time.Now the base.sync package tests everything using the TaskService interface. Right now there are not the same tests for the sync and theasync interface but I can migrate them soon.
2) Reviewing the org.jbpm.task.service package I found all the UserGroupCallback tests and the class called BaseTestNoUserGroupSetup, which basically run almostall the test using no hardcoded users and using the UserCallbacks resolutions. Talking yesterday with Tiho, we both agree that we should remove the teststhat are using hardcoded users and just leave the one that are using callbacks to avoid running unnecessary tests. Just doing this we will remove a lot of overheadand we can have a simplified way of organizing and running the tests. I've moved all the UserGroupCallback tests to the async package and migrate them to use the AsyncTaskService instead of the TaskClient.
3) I've created a package called org.jbpm.task.identity to place the LDAP stuff which is ignored and the UserGroupCallBack stuff that is not using the TaskService/TaskClient stuff, when I was doing that I notice that those tests uses some properties files that needs to be placed inside the same package structure to be picked up, for that reason I've moved two classes inside src/main/java to the following new package: org.jbpm.task.identity -> UserGroupCallbackManager, UserGroupCallback, DefaultUserGroupCallbackImpl, LDAPUserGroupCallbackImpl, which I believe that is better than have the UserGroupCallback stuff mixed with the task.service stuff. I notice that the package: org.jbpm.task.service is a wild area, that we should avoid as much as possible. I know that this is a change that can impact in other modules like the jbpm-console but I think that it will avoid future troubles. I really need your feedback on this.. to see if you guys think that this can cause a lot of troubles.
4) for the same reason as 1 I've commented out all the tests inside org.jbpm.task.service.test, those were using the TaskClient and because the async and sync test are already provided we don't miss anything. The only thing that we need to do inside the org.jbpm.task.service.test package is to align these tests to cover all the base async and sync tests, which is not the case right now, but it's simple to do it.
5) I've added covertura to the jbpm-human-task-core project to do an initial analysis about what we are covering with our tests. Attached is the initial chart that I would love to see improved in the following months :)
6) I need to work on the other modules to propagate these changes as well, but for doing that I need your feedback on the previous points, I think that during the weekend I will be able to finish the other modules, if we all are OK with this suggestions.
Cheers
--
- MyJourney @ http://salaboy.wordpress.com- Co-Founder @ http://www.jugargentina.org
- Co-Founder @ http://www.jbug.com.ar
- Salatino "Salaboy" Mauricio -
-- jBPM/Drools developer Utrecht, the Netherlands
--
- MyJourney @ http://salaboy.wordpress.com- Co-Founder @ http://www.jugargentina.org
- Co-Founder @ http://www.jbug.com.ar
- Salatino "Salaboy" Mauricio -
_______________________________________________
jbpm-dev mailing list
jbpm-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbpm-dev
-- jBPM/Drools developer Utrecht, the Netherlands