On Fri, May 3, 2013 at 9:25 AM, Thijs Kinkhorst wrote: > On Fri, May 3, 2013 15:09, Wouter Verhelst wrote: >>> > No, it's not. Source only uploads were banned many years ago, mainly >>> due >>> > to problems with maintainers not even build testing their packages. > >> They do. They just ignore the issue; they can do that because it's a >> scalability issue that, ultimately, can be fixed by throwing large >> amounts of hardware (for more build hosts) at it. Debian does not have >> that luxury. > > I'm not sure how we can know the amount of unneeded build failures would > be caused by such a strategy. Why do you think the current infrastructure > can't handle these cases? How many do you think there will be, as a > percentage?
I wonder if a somewhat straightforward solution would be for dak to delay its usual processing (for source-only uploads only) until the buildds produce binary packages on at least one arch? Then, in the case of a build failure on all archs, the upload would of course be automatically rejected. This kind of approach would at least keep fully unbuildable packages out of unstable (it doesn't prevent arch-specific build failures), which is what I think people really want. This of course would require some additional communication paths between dak and the buildds. Best wishes, Mike -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/CANTw=MPyq=f=Bw2-2K8_ZnQroOGA851N3g-r=h2V=gwtsne...@mail.gmail.com
