Hello Clement, On Mar 26, 2010, at 9:29 , Clement Escoffier wrote:
> On 26.03.2010, at 17:23, Marcel Offermans wrote: > >> On Mar 26, 2010, at 8:19 , Guillaume Nodet wrote: >>> On Fri, Mar 26, 2010 at 15:58, Sahoo <sa...@sun.com> wrote: >>>> Pierre De Rop wrote: >>>>> However, the -Dmaven.test.skip=true option still has to be used because >>>>> there is one pending issue in the DependencyManager junit tests, which has >>>>> to be fixed. This is a known issue, and it's referenced here: >>>>> http://issues.apache.org/jira/browse/FELIX-2078. >>>> >>>> >>>> Why don't we exclude the test until it is fixed? That will keep CI server >>>> to function as well. >>>> >>> >> >>> That's a good idea ;-) >> >> Whilst I don't mind too much doing that, I'm wondering why our build cannot >> support subprojects that have failing tests? I mean as long as I'm not >> releasing anything, a project that is in flux might have a few test cases >> that are broken, right? That should not affect our whole build in a way that >> it makes life hard on everybody else. >> >> Is it really not possible to have a solution where having tests that fail is >> not a problem? >> >> If not, then just disable that test for now. > > There is several things that are possible: > - setting 'testFailureIgnore=true' ignore test failures (not sure about > errors). I'm no expert on what this option does, but it sounds like a solution. > - executing the tests only when a 'test' profile is enabled. That was my > choice for iPOJO because of the number of tests. That's certainly an option, so the normal build someone does simply builds the artifacts without any tests whatsoever, and on the CI servers we run this test profile that enables all of them (probably combined with that option you mention so the build does continue even if it encounters test failures). Does this make sense (to the more Maven experienced amongst us)? Greetings, Marcel
