Re: [gentoo-portage-dev] [PATCH] use.force and package.use.force (bug #142853)

[Brian Harring]

On Mon, Aug 07, 2006 at 03:52:35PM +0000, Alec Warner wrote:
>
> >> Brian, default USE in IUSE is not a backwards compatable change and this
> >> is easier ;)
> >
> > An EAPI bump is pretty simple from where I'm sitting, and implementing
> > it isn't all that hard.
> >
> > Meanwhile, the question should be "which is desirable", not "which can
> > I glue in quickest" ;)
> >
> > ~harring
>
> This is reality, not the ideal;

Was also what devs wanted, last time this was discussed on dev ml from 
what I recall.

Meanwhile, this is not an 'ideal' if you think it through.

The request is effectively per node use configuration mangling of the 
default state- reality being that package.use _already_ exists, and 
shoehorning which ever is implemented relies on the same bits.

> and implementing a dep resolver like you
> wanted IS hard, thats why it took you and stubbs a ton of time
> figuring out what you wanted and then implementing it.

Way off base here.  See the paragraph above- further, while this 
involves the resolver, it's not actually a resolver mod.

<sidenote>
The design jstubbs and I kicked around (what is now in pkgcore) goes 
several steps beyond what actually is needed- said design break config 
instances down and binds it per pkg instance- that's more of an issue 
for supporting actual use deps though, although said design isn't a 
requirement; paludis supports it via using (effectively) a construct 
similar to portages config class.

We (moreso me on this one) wanted a format agnostic resolver- the 
design/implementation reflects that choice, but doesn't mean that 
route is the only way (again, paludis is closer to portage in this 
regard- it's doable).
</sidenote>

Meanwhile, we're talking about default iuse here; totally unrelated 
subjects.

Resolver actually doesn't have a helluva lot to do with this in 
portages design- at the very least the view that default IUSE requires 
a "dep resolver like you wanted" (bit vague) pretty much has no 
relevance, and indicates you're a bit confused on this (especially 
since the critique is leveled at default IUSE while ignoring that 
use.force is no different).

See zacs patch- both use.force and default IUSE require going back 
to the config obj at some point- the difference is contained in 
config; finally, EAPI protection should work fine for it- prefix has 
been relying on that already for a long ass time now.

~harring