On Thu, 14 Jul 2016, Emmanuel Bourg wrote: > Le 14/07/2016 à 14:08, Santiago Vila a écrit : > > > Would you apply the same reasoning to a package which fails to > > build from source in every arch-dependent autobuilder but still builds > > fine without the -B flag? > > No I wouldn't but that's a different case. > > If I build on amd64 and do a binary upload I can't ignore build failures > with -B, because the package won't be available on the other > architectures and that's indeed a serious issue impacting the users. On > the other hand, if -A fails I'm still able to upload the architecture > independent packages and there is no impact for the users. > > -A failures are only relevant for source only uploads, and as long as > they aren't mandatory I don't think this kind of issue should have a > severity higher than important.
What you call "binary upload" is mostly an upload which includes all the "Arch: all" binary packages. If you were able to provide all the "Arch: any" binary packages, failure to build in the official autobuilders would not be serious, then? Consider a package which is both Linux and Intel specific, only buildable in amd64 and i386. With a little bit of trickery, you could make an upload including all the required packages even if they fail in the official autobuilders. Would this be ok? I don't think so. Or even better: Consider a source package which only builds "Arch: all" binary packages. Suppose your package has a missing build-depends but it builds ok in your computer because you have installed the build-dependency in your chroot. Since you provide all the "Arch: all" packages, would it be ok not to consider the FTBFS in an official autobuilder as serious? I don't think so either. Thanks. __ This is the maintainer address of Debian's Java team <http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-java-maintainers>. Please use [email protected] for discussions and questions.
