On Mon, Oct 20, 2003 at 10:24:19AM +0200, Bill Allombert wrote:
> On Mon, Oct 20, 2003 at 12:15:17AM -0500, Chris Cheney wrote:
> > On Sun, Oct 19, 2003 at 08:15:43PM +0100, Andrew Suffield wrote:
> > > On Sun, Oct 19, 2003 at 01:20:39PM +0100, Scott James Remnant wrote:
> > > Other people have covered why this breaks. Here's the solution I use:
> > >
> > > Make your build target do nothing.
> > >
> > > That is, make build an empty target that does _not_ depend on
> > > build-arch and build-indep. Then make sure that binary-{arch,indep}
> > > will result in the right things getting run anyway.
> > >
> > > This a) preserves the desired effect of the time consuming arch-indep
> > > stuff not being run on the buildds, and b) actually works. While it's
> > > not strictly in accord with policy as written, it has the advantage of
> > > doing what policy expected to happen, and I've never seen a better
> > > idea.
>
> There is a better idea that presever point a) and b) above:
> write dependencies as follows:
> build: build-arch
> binary-indep: build-indep
>
> This work since policy require both Build-Depends and
> Build-Depends-Indep to be satistied when running build-indep.
>
> The defect is that build-indep is made as root. The advantage over
> Andrew solution is to not make build-arch as root.
We haven't built packages as root for years. This argument seems
pretty irrelevant.
> > > Ultimately we should either deprecate the build* targets, or make
> > > build-{arch,indep} mandatory and deprecate build.
>
> Again there are other possibilities, like using make variable for the
> transition.
I say this is impossible, on account of it being precisely equivalent
to using the build* targets. Describe such a possibility in detail if
you think otherwise.
--
.''`. ** Debian GNU/Linux ** | Andrew Suffield
: :' : http://www.debian.org/ |
`. `' |
`- -><- |
signature.asc
Description: Digital signature
