Hi, Le lundi 17 mai 2021 à 15:20 +0200, Jochen Sprickerhof a écrit : > Hi Julien, > > * Julien Puydt <julien.pu...@gmail.com> [2021-05-17 09:01]: > > I tried to create a testing sbuild and compile sagemath 9.2-2 with > > it, > > and it worked so unless I made a mistake in my setup, something > > made > > that bug go away... > > > > Can someone independently give it a try? > > I triggered reproducible-builds again: > > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/sagemath.html > > Success: 40 tests failed, up to 200 failures are tolerated > Success: 5 tests failed, up to 200 failures are tolerated > > so not much changed comparing to two weeks ago and my conclusion > still > holds: > > * Jochen Sprickerhof <jspri...@debian.org> [2021-05-04 13:22]: > > Success: 41 tests failed, up to 200 failures are tolerated > > Success: 5 tests failed, up to 200 failures are tolerated > > > > The 200 is set in debian/rules: > > > > https://salsa.debian.org/science-team/sagemath/-/commit/6088e9f2abc7ae99a8d07760ceee6fb6aac7bb54 > > > > and sounds a little arbitrary. Sadly the state upstream seems not > > to > > be much better: > > > > https://github.com/sagemath/sage/tree/9.2 > > > > 13 failing, 17 cancelled, and 70 successful checks > > > > (I did not look into them.) > > > > So I think the question is rather if the test suite gives an > > appropriate view on the quality of the software. If it does, I > > assume > > it is not appropriate for a Debian stable release. Contrary if we > > assume the test suite being broken, we could disable it completely > > rather then producing random FTBFS.
Well : 1) Upstream itself uses the testsuite in the sense of "shouldn't have too many failing tests", and it still allows to detect when a build is utterly broken, so we shouldn't disable it. 2) It's not an FTBFS, since the sources actually build to a set of packages. This points to lowering this bug severity and let sagemath go in stable. Or at least not preventing it just for this reason. JP