Re: [gentoo-dev] Sets in the tree

[Michał Górny]

Dnia 2013-08-15, o godz. 11:09:50
Tom Wijsman <TomWij@g.o> napisał(a):

> On Thu, 15 Aug 2013 10:10:02 +0200
> Pacho Ramos <pacho@g.o> wrote:
>
> > El mié, 14-08-2013 a las 23:53 +0800, Patrick Lauer escribió:
> > [...]
> > > Well, it should reflect reality.
> > >
> > > PMS is still broken as much as it does not reflect the state of
> > > portage before PMS was written, and we've had to patch it up a few
> > > times to make it coherent, plus it is still lacking half the things
> > > that would make it useful as a standard.
> > >
> > > Your academic interpretation of standard as a platonic ideal
> > > disconnected from reality serves no purpose.
> > >
> >
> > On this topic I agree with Patrick: I don't fully understand why
> > things (like in_iuse from eutils.eclass) are missing from PMS.
>
> Why do things that can be in an eclass need to be in the PMS?
>
> Or more specifically, why would you like to see in_iuse in the PMS?

in_iuse() is a cheap hack that does not confirm to the PMS.

The problem is that PMS provided us with no way to query incremental
variables like $IUSE. It makes the value of such variables 'undefined'.
Which, shortly saying, means something like 'you can't query it'.

Therefore, in_iuse() is broken and unsupported and you will be damned
for using it. The only reason I added it to eutils.eclass was because
people did it in even a more broken way. It's mostly like 'we know it's
broken, so better keep it confined'.

What future EAPI could do is to provide a legitimate way of querying
it. Not that I really agree with doing that but it's not my call how
people mis-design eclasses, and that's out-of-topic here.

-- 
Best regards,
Michał Górny