Op do 23-10-2003, om 22:03 schreef Nick Phillips: > On Wed, Oct 22, 2003 at 05:00:34PM +0100, Andrew Suffield wrote: > > On Wed, Oct 22, 2003 at 10:02:06AM -0500, John Hasler wrote: > > > Andrew Suffield wrote: > > > > I say that failing to function when built in anything but a particular > > > > artificial environment is a serious bug in a source package. > > > > > > The environment on the workstation of any random developer is no less > > > artificial than of the autobuilders. The latter, however, is consistent, > > > reproducible, and designed to minimize variables. > > > > And the whole point is that this is not desireable. > > > > Why has nobody been paying attention? > > I think you have now posted at least a couple of messages which successfully > explain your point. > > You think that having tested a package which was built in a "real-world" > situation is critical. ... and I agree with Andrew. > Most people who are arguing with you think that it is more important that > the binary packages we ship are built in a know, reproducible, and consistent > way. > > If we keep the status quo as you suggest, we will be hit by the occasional > case where something goes wrong on autobuilt packages because the maintainer > accidentally did something "special" (which could even be part of Debian) on > his machine when building. We also get the "HTF did this ever build at all?" > bugs. Yes, and we wouldn't catch them if everything was autobuilt for every architecture (your point may be that those bugs wouldn't exist if we didn't autobuild for every architecture; that's incorrect, it simply would fail on one extra buildd) > If we autobuild everything, we may get the reverse; the occasional problem > that only occurs when you have something "special" (which could be part of > Debian) on the build system. Yes, but we wouldn't catch those bugs. > Either way, occasionally you will get nasty problems that the other method > would have caught. > > The difference is that autobuilding everything is reproducible, consistent, > predictable and all those other things that good engineering practice is > supposed to be. We _do_ autobuild everything. Just not on every architecture. > Eliminating sources of human error is GOOD. I can't agree more with this phrase. -- Wouter Verhelst Debian GNU/Linux -- http://www.debian.org Nederlandstalige Linux-documentatie -- http://nl.linux.org If you're running Microsoft Windows, either scan your computer on viruses, or stop wasting my bandwith and remove me from your addressbook. *now*.
Attachment:
signature.asc
Description: Dit berichtdeel is digitaal ondertekend