Re: force-confnew (was: Document correct buildd chroot setup somewhere?)
On Mon, Apr 05, 2010 at 02:23:49PM +0200, Andreas Barth wrote:
> * Kurt Roeckx (email@example.com) [100405 14:19]:
> > On Mon, Apr 05, 2010 at 01:58:31PM +0200, Andreas Barth wrote:
> > > However, I still think we should use force-confnew, because conffiles
> > > shouldn't contain valuable changes in the chroots. Why do you think
> > > it's bad to use force-confnew?
> > It hides bugs.
> and? I think these are bugs that we should hide at least in
> [old]stable and on slow architectures. That's not the only place we
> don't fall on bugs.
For instance we now run into #576399. And I'm not sure the config
file from libao2 will work with libao4, or the other way around.
The package might fail to build for some other reason that might
not be as clear.
> (E.g. the we don't uninstall packages hides bugs as well. But I'm
> happy with that.)
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.
> An alternative suggestion: How if we grep for the specific occurence
> in the build logs, and file bug(s) (or even autofile once per week
> against buildd.d.o if there was at least one occurence)? That way we
> don't block builds, but we still notice.
There are various issues we can grep for in buildds logs, and I
would like it that someone would do something like that.