El Divendres, 18 de gener de 2013, a les 18:58:07, Martin Gräßlin va escriure: > On Tuesday 15 January 2013 23:27:53 Albert Astals Cid wrote: > > 4) Don't release if any if the tests are failing in builds.kde.org > > > > If we have tests, they have to work > > I have mixed feelings about that. Personally I agree completely, but I fear > that this will end in less unit tests than in more. If you have unit tests > and they have been broken for a long time you get punished,
Right, but why would you have a broken unit test? The only reason i can think of is that you know your code is broken but don't have time, etc to fix it yet. If that's the case i'd suggest using QEXPECT_FAIL maybe? > if you don't > have tests, your code is accepted without any problems. So it kind of > encourages to have bad code (no unit tests). > > Before we go this way I would suggest to first get a cleanup on the unit > tests and enable nag mails when a test starts to break [1]. If we see that > this works well, we could go the full thing in 4.12. Also I would like to > see some way to encourage more unit testing. Sure, I'm not saying how we should implement this, just asking if people agree that "tests should pass" :D Of course we should either kill or comment or EXPECT_FAIL the tests. Cheers, Albert > > Cheers > Martin > > [1] I would love to get those, as I cannot use the build.kde.org build > status for kde-workspace due to three tests not in KWin failing all the > time _______________________________________________ release-team mailing list [email protected] https://mail.kde.org/mailman/listinfo/release-team
