On 12/07/2012 05:33 PM, Thiago Macieira wrote: > On sexta-feira, 7 de dezembro de 2012 17.12.06, Jedrzej Nowacki wrote: >>> I agree in theory, provided that the tests that do stress out the >>> integration are close to complete testing. In practice, there's no way to >>> tell whether the testing is enough, so I'd rather we keep full testing, >>> whether redundant or not, unless we have a compelling reason not to. >> >> I think that "being able to run all tests on a desktop without focus issues" >> + "reducing test time execution from 8,5 to 2 minutes" = "a compelling >> reason". > > I don't think that's enough a risk of introducing regressions because we don't > have any testing of a particular feature. Note I was talking about running in > the CI system, not for quick tests locally before submitting.
Agreed, we don't want to lose any valuable coverage in the process. If we did make a QtWidgets test run offscreen in the CI we would have to ensure that we made up for any loss of coverage in QtGui auto tests. > A "compelling reason" for running in the CI system would be that the total > test time is too long, for example. Yep, speeding up CI would pretty much be the only reason. -- Samuel _______________________________________________ Development mailing list Development@qt-project.org http://lists.qt-project.org/mailman/listinfo/development