Re: force-confnew (was: Document correct buildd chroot setup somewhere?)
* Philipp Kern (email@example.com) [100405 20:36]:
> On Mon, Apr 05, 2010 at 02:47:40PM +0200, Kurt Roeckx wrote:
> > And I think it's a good thing that not all buildds use lvm chroots
> > so that we actually find problems we might not see otherwise, or
> > maybe much later.
> Sadly, I have to differ on this one. I'd rather appreciate a coherent,
> identical setup everywhere where it's possible. Either fail everywhere
> due to that issue or nowhere. Having setup differences that boil down
> to "this buildd of that arch behaves differently than the others" or
> "I guess this is arch-specific because it only happens there" (while
> it's just due to the setup), does not help IMO.
I think it might be a good idea to also uninstall packages on a few
buildds, but without failing the builds but with appropriate messages
in the build log. This comes without real cost on fast arches, and
doesn't really have impact (it is *after* the builds, and sbuild
should write something like "errors after this line are ignored").
And, we discussed that earlier,
confnew: If a conffile has been modified always install the new
version without prompting. (instead of prompting as usual)
whereas it doesn't steal a file from another package without replacing
overwrite: Overwrite one package's file with another's file.
and I agree that this shouldn't be set), just to avoid confusion.