On Wed, Dec 10, 2008 at 11:08:27AM -0600, Shawn Walker wrote:
> Bart Smaalders wrote:
> > Tom Mueller (pkg-discuss) wrote:
> >> In the facet and variants design, would it make sense to allow a package
> >> to declare a dependency on the image having a certain facet? This would
> >> allow packages containing compilers, Dtrace, etc. to declare that
> >> headers need to be installed.
> >
> > I was mulling over something like this as well... I'd rather do this
> > than to have implicit dependencies ("make sure to grep through the
> > package list and install everything that's -devel to pick up the
> > headers"). I'm still mulling over the user experience; the side
> > effect of installing additional facets of already installed packages
> > seems pretty analogous to installing additional dependencies.
> >
> > In general, this will somewhat hinder minimization efforts, but that's
> > prob. ok here.
> >
> > What it argues for is defining a sufficient number of facets to prevent
> > the installation of large numbers of unneeded files.
>
> That pretty much sums up the idea that I was trying to get across, but
> apparently articulated poorly.
I agree with the mechanism.
But again, it seems likely that DTrace will generally be installed
(doesn't it have to be? doesn't Java depend on it?), and most uses of
DTrace don't require headers to be installed (nor cpp). I'm still not
sure when headers should be installed now that there exist apps (e.g.,
some DTrace apps) that need them at runtime.
> Having the ability for packages (such as DTrace or a compiler) to say
> that they needed the "headers" facet installed would be a boon, since it
> would allow headers to be installed only if there was software that
> needed them.
Yes.
Nico
--
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
