On Sun, Feb 11, 2007 at 12:05:07AM +0100, Axel Thimm wrote:
On Sat, Feb 10, 2007 at 10:41:44AM -0800, Toshio Kuratomi wrote:
> On Sat, 2007-02-10 at 14:53 +0100, Axel Thimm wrote:
> > e) True arguments are that
> > obscuring the buildroot for the sake of an extremely rare
> > corner case (several users building the same package on the
> > same system w/o chroots) implies fixing it for far more not
> > corner-cases like building i386 and x86_64 packages
> > simulataneously. So the `id -nu` part is far less important
> > than adding the target arch, but that was silently forgotten by
> > racor
> >
> Is your suggestion to add arch to the buildroot?
No, my suggestion is to loose up on requirements on buildroots. There
is no known problem ever caused by choosing a "wrong" buildroot even
by novices, and we're definitely over-engineering in fixing stuff that
never broke.
I completely agree with that. Unless the buildroot is picked by
mkdtemp() you can't really *guarantee* avoidance of conflicts. If you
want a guarantee then rpmbuild should be fixed to ignore BuildRoot and
use mkdtemp() instead. Standardising an inadequate workaround and
having packagers go through fixing N hundred spec files to match seems
like a waste of time.
joe