+1 For me it is better to introduce a new dependency if we can create better tests (and faster) than using nasty workarounds...
Bruno P.S. Now in the correct place :-) 2005/7/21, Bill Dudney <[EMAIL PROTECTED]>: > Hi All, > > I've got the first of the EasyMock (http://www.easymock.org) tests in > place. I have not commited the changes because I wanted to get > everyone's opinion before introducing a new dependency. There are 4 > new jars required to use easymock, ams.jar, cglib.jar, easymock- > classextensions.jar and easymock.jar. The cglib and ams dependency > are used to mock abstract classes (the default easymock.jar is able > to mock interfaces without cglib or ams). > > With the EasyMock framework I was able to get to 100% code coverage > of the abstract StateManager class with 5 tests and less than 125 > lines of actual test code. I'm a fan and have used it extensively on > other projects. I believe the extra dependencies are worth it to get > the testing done. > > As soon as there is consensus I'll commit the changes to build.xml > and the actual test code (or by tomorrow afternoon or so assuming > that a lack of comment is agreement) then move on to the cactus test > stuff. > > Sean: I'll try a couple of different approaches to the build and post > in the form of a proposal, so your feedback to that thread will be > most useful :-) > > TTFN, > > -bd- > > On Jul 20, 2005, at 1:39 PM, Bill Dudney wrote: > > > Hi Grant, > > > > Thanks for your interest in the testing stuff. > > > > Sorry I've not committed the cactus stuff yet. I need to get my > > thoughts together in an new thread that Sean can comment on because > > there will be additional build stuff in place (cargo to start/stop > > containers, packaging a new war file to distribute the cactus tests > > in etc.) and I want the changes to fit with Sean's continued vision > > of the build process. > > > > In the mean time we could use mock objects. Although the class you > > refer to (ServletFacesContextImpl) is particularly hard to test > > because the lack of default config setup that happens during out of > > container testing. > > > > I have a test for the Factory that could be used as the starting > > point of a factory setup so that the rest of the required stuff > > could be mocked for the facescontext impl. I will get that checked > > in asap (SVN appears to be down again). > > > > TTFN, > > > > -bd- > > > > On Jul 20, 2005, at 11:36 AM, Grant Smith wrote: > > > > > >> Bill, > >> > >> The current testing setup is great for objects that are container- > >> independent. Any indication when we'll be able to test things that > >> need to live in a container, like FacesContextImpl ? I assume > >> we'll need cactus for that, although if you know of any tricks to > >> test container-dependent objects, that would be valuable > >> information :) > >> > >> Another question for cactus gurus: is cactus portlet friendly ? > >> > >> Thanks, > >> Grant > >> > >> > > > > > >
