On Thu, Mar 10, 2011 at 07:57:44PM +0000, Philipp Kern wrote: > On 2011-03-10, Bernd Zeimetz <firstname.lastname@example.org> wrote: > > On 03/10/2011 06:56 PM, Philipp Kern wrote: > >> On 2011-03-10, Hector Oron <email@example.com> wrote: > >>> I was planning to start using a wanna-build instance, but i was not > >>> sure which one to use. > >>> As I have not yet started to use any of them I have no real objection > >>> for its removal, on the other hand I would not mind to try to merge > >>> current wb used on buildd with the one shipped in sbuild. Just let me > >>> know your preferences. > >> The two codebases don't have much in common, given that the sbuild one was > >> heavily refactored. > > So why the hell are these two branches not merged and developed by a team? > > Volunteer time is limited and the focus of the wanna-build team is a different > one than providing packages, at least at the moment. (Especially as it's > hard to produce something that works equally well as a package and as a > central installation that's updated differently.) There's also no benefit > in moving the central wanna-build to a refactored code base that has less > features and fixes. Exactly. I was planning to merge the wanna-build.git changes into sbuild.git for a *long* time. I just haven't had time to do it, and no one else has seemed especially interested. Certainly not sufficiently interested to volunteer their time to do it. > buildd in unstable was severely broken for ages, because the packages > weren't at all tested after heavy refactorings (and still aren't > because the maintainer doesn't have a test setup). wanna-build will > have suffered the same fate. The main sticking point for me being able to properly test buildd is that is requires a functional wanna-build setup to do so. And setting one up is decidedly non-trivial. wanna-build.git is unusable without all the extra support scripts on grieg, and the schema files are broken. I was not able to set up a working wanna-build environment. What I'm thinking of doing here is writing a "fake" wanna-build that can emulate the real one. To the extent that e.g. --list=needs-build can return a list of packages to build. This would allow buildd testing in the absence of a full wanna-build setup. Again, it needs time to do. Any help here would be appreciated. wanna-build was a little different from buildd. It was /never/ truly functional from the start, either before or after any refactoring, and with zero users, there just hasn't been the demand to work on it above sbuild or buildd. > For buildd-0.61.0 the situation seems to be a bit better, I only needed > some few hours instead of days to get it working again. buildd *is* living > in the same repository as sbuild, though. It's just wanna-build that's > separate. That's good to know. Adding the testsuite has reduced the chance of regressions somewhat. Getting buildd in the testsuite as well would be even better. > That said, if you want to step up in maintaining a hellish Perl codebase, > feel free to send patches. Help is always appreciated. It's not the nicest codebase in the world, but I think we've improved it significantly in recent years. sbuild at least is now usable by mortals, though it could still be easier to set up; documentation is the main lacking here. Regards, Roger -- .''`. Roger Leigh : :' : Debian GNU/Linux http://people.debian.org/~rleigh/ `. `' Printing on GNU/Linux? http://gutenprint.sourceforge.net/ `- GPG Public Key: 0x25BFB848 Please GPG sign your mail.
Description: Digital signature