> On Jan. 23, 2015, 8:43 a.m., Vishesh Handa wrote: > > I'm not against this, but I am curious as to why this is being done. > > > > I would think that packagers should be building the tests and running them > > on their platform and make sure everything passes. We have a strict policy > > that all tests must always pass. > > Matthew Dawson wrote: > This is mostly useful on source based distributions (specifically, this > patch comes from Gentoo). While in general running tests everywhere would be > great, source distro users may not have the cpu time to compile/run tests. > Also, some test suites don't work and users may not care to figure out why > (for instance, last time I tried enabling tests in Gentoo, binutils failed > its suite). > > For binary distriubtions, I agree they should be running tests > (especially since we work to keep them green). But source based distros > aren't so clear cut. > > Andreas Sturmlechner wrote: > Exactly, packagers do build the tests of course, but that does not mean > users of source packages should have a permanent dependency on Qt5Test. > > Vishesh Handa wrote: > It is even more important for source based distros to be running tests. > They generally have very different compile options and flags. What is the > point of them running the software and possibly finding bugs, when it could > have been caught by just running the tests. > > Actually, the more I think about this, the more I realize that everyone > should be running the autotests. -2 from my side. But I'm not the maintainer > of kio. > > Vishesh Handa wrote: > > Exactly, packagers do build the tests of course, but that does not mean > users of source packages should have a permanent dependency on Qt5Test. > > On a source based distribution you already have a dependnecy on cmake, > the compiler, and many other things. These are only required during build > time. Qt5Test is the same. Once the pacakge has been built + tests have been > run, Qt5Test can be removed. > > Matthew Dawson wrote: > At least on Gentoo, by default build time dependencies are not > automatically removed (though you can remove them if you want). Generally > speaking that is the right choice, as you will need cmake/compiler/etc. later > to build the package when a version is released. Also, one of the benefits > of Gentoo is that the entire development toolchain sticks around, allowing > for easy development/bug triaging. > > Anyways, source based distros won't always run tests, because users won't > always want to run them. In a perfect world, I agree that is wrong. In > reality, I don't run any test suites across any of my Gentoo installs. So > forcing the tests to build just burns CPU time, and is easily patched out by > downstreams. I don't think trying to force this will get KDE anywhere. > > Vishesh Handa wrote: > > Anyways, source based distros won't always run tests, because users > won't always want to run them. In a perfect world, I agree that is wrong. > In reality, I don't run any test suites across any of my Gentoo installs. So > forcing the tests to build just burns CPU time, and is easily patched out by > downstreams. I don't think trying to force this will get KDE anywhere. > > - If the user doesn't want to run them, I'm sure Gentoo can provide some > options for that. Compiling them cannot be such a huge cost. > - They are already burning CPU time by using a source based distro. This > way they might actually catch some bugs and possibly not waste developers > time by filing bugs which may have been an issue with their system. > > I'm not sure if I will approve such patches on packages I maintain. > > Matthew Dawson wrote: > I think we've both stated our piece here, and we aren't going to get > further towards a consensus. May I suggest posting this to the general > kde-frameworks (or kde-core-devel, I'm not sure what would be better) seeking > to make a general policy wrt Frameworks? That way Frameworks has a > consistent approach to building tests, whatever way the community decides. > > In the meantime we should probably merge this patch, as building > autotests without finding Qt5Test is only going to break builds. Then > packages can be updated with the policy decision by removing the > BUILD_TESTING option. > > The Plasma community should also probably come to a consensus for its > packages as well. > > Albert Astals Cid wrote: > Your logic is flawed, you say "building autotests without finding > Qt5Test is only going to break builds". > > That is correct, just that Qt5Test is being searched for, so your > rationale for applying the patch is moot.
Sorry, I misread the patch. You are correct the build is fine as is. We should discuss this on the mailing list first before applying. - Matthew ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://git.reviewboard.kde.org/r/122206/#review74602 ----------------------------------------------------------- On Jan. 22, 2015, 2:48 p.m., Andreas Sturmlechner wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://git.reviewboard.kde.org/r/122206/ > ----------------------------------------------------------- > > (Updated Jan. 22, 2015, 2:48 p.m.) > > > Review request for KDE Frameworks. > > > Repository: kio > > > Description > ------- > > [kio] Make tests optional > This is a small patch to CMakeLists.txt to only depend on Qt5Test if > BUILD_TESTING. > > > Diffs > ----- > > CMakeLists.txt 7fe0be5d4b2d7d9475a7844b4f8d93fc2f0a00c3 > > Diff: https://git.reviewboard.kde.org/r/122206/diff/ > > > Testing > ------- > > > Thanks, > > Andreas Sturmlechner > >
_______________________________________________ Kde-frameworks-devel mailing list Kde-frameworks-devel@kde.org https://mail.kde.org/mailman/listinfo/kde-frameworks-devel