Mailing List Archive

RFC: making USE_EXPANDed variables incremental
Following a discussion in #gentoo-portage earlier this evening, it was
suggested that I send out an RFC email for this. So, does anyone object
to requiring that any variable listed in USE_EXPAND be treated as
incremental, at least as far as profile inheritance is concerned?
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
Stephen Bennett wrote:
> Following a discussion in #gentoo-portage earlier this evening, it was
> suggested that I send out an RFC email for this. So, does anyone object
> to requiring that any variable listed in USE_EXPAND be treated as
> incremental, at least as far as profile inheritance is concerned?

That means that the base profiles must have a minimal setting that is
added to in lower profiles, rather than a reasonable default that's
entirely reset in lower profiles (perhaps to a smaller setting), correct?

I'm not a huge fan of that, if that's what it requires, since there is
no way of subtracting USE_EXPAND settings that I know about.

Thanks,
Donnie
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
Donnie Berkholz wrote:
> I'm not a huge fan of that, if that's what it requires, since there is
> no way of subtracting USE_EXPAND settings that I know about.

Ability to subtract comes with incremental stacking, which I think we
are talking about here, at least I hope so.

--
Kind Regards,

Simon Stelling
Gentoo/AMD64 Developer
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Mon, 01 Jan 2007 13:24:49 -0800
Donnie Berkholz <dberkholz@gentoo.org> wrote:

> That means that the base profiles must have a minimal setting that is
> added to in lower profiles, rather than a reasonable default that's
> entirely reset in lower profiles (perhaps to a smaller setting),
> correct?

It would mean that all USE_EXPANDed variables get stacked in the same
way that USE does. The base profile defines a set of defaults, which
gets flags added to or removed from it in other profiles. At present,
from what zmedico told me, it's handled in a weird manner which
essentially does half the job, letting you add flags but not remove
them in subprofiles.
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
Stephen Bennett wrote:
> On Mon, 01 Jan 2007 13:24:49 -0800
> Donnie Berkholz <dberkholz@gentoo.org> wrote:
>
>> That means that the base profiles must have a minimal setting that is
>> added to in lower profiles, rather than a reasonable default that's
>> entirely reset in lower profiles (perhaps to a smaller setting),
>> correct?
>
> It would mean that all USE_EXPANDed variables get stacked in the same
> way that USE does. The base profile defines a set of defaults, which
> gets flags added to or removed from it in other profiles. At present,
> from what zmedico told me, it's handled in a weird manner which
> essentially does half the job, letting you add flags but not remove
> them in subprofiles.

I'd rather not make USE_EXPAND incremental if we can't subtract flags.
At present, we accomplish that by simply resetting the whole thing in
subprofiles. But the proposal seems to make impossible any subprofile of
a valid profile that wishes to negate a setting of the parent.

Thanks,
Donnie
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Mon, 01 Jan 2007 15:27:43 -0800 Donnie Berkholz
<dberkholz@gentoo.org> wrote:
| I'd rather not make USE_EXPAND incremental if we can't subtract flags.

You mean via -flag? Or via -*? Both are valid in incrementals.

--
Ciaran McCreesh
Mail : ciaranm at ciaranm.org
Web : http://ciaranm.org/
Paludis is faster : http://ciaranm.org/show_post.pl?post_id=61
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Mon, 01 Jan 2007 15:27:43 -0800
Donnie Berkholz <dberkholz@gentoo.org> wrote:

> I'd rather not make USE_EXPAND incremental if we can't subtract flags.
> At present, we accomplish that by simply resetting the whole thing in
> subprofiles. But the proposal seems to make impossible any subprofile
> of a valid profile that wishes to negate a setting of the parent.

I wrote:
> > It would mean that all USE_EXPANDed variables get stacked in the
> > same way that USE does. The base profile defines a set of defaults,
> > which gets flags added to or removed from it in other profiles.

The proposal means that all variables listed in USE_EXPAND get handled
exactly as USE does where profile inheritance is concerned. Subprofiles
can add to and remove from the value in the parent profile just as they
can for USE.
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
Stephen Bennett wrote:
> The proposal means that all variables listed in USE_EXPAND get handled
> exactly as USE does where profile inheritance is concerned. Subprofiles
> can add to and remove from the value in the parent profile just as they
> can for USE.

Did I misread what you said earlier?

Stephen Bennett wrote:
> At present,
> from what zmedico told me, it's handled in a weird manner which
> essentially does half the job, letting you add flags but not remove
> them in subprofiles.

Thanks,
Donnie
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Mon, 01 Jan 2007 17:24:03 -0800
Donnie Berkholz <dberkholz@gentoo.org> wrote:

> Stephen Bennett wrote:
> > The proposal means that all variables listed in USE_EXPAND get
> > handled exactly as USE does where profile inheritance is concerned.
> > Subprofiles can add to and remove from the value in the parent
> > profile just as they can for USE.
>
> Did I misread what you said earlier?
>
> Stephen Bennett wrote:
> > At present,
> > from what zmedico told me, it's handled in a weird manner which
> > essentially does half the job, letting you add flags but not remove
> > them in subprofiles.

"At present" -- that's the behaviour that I want to change by making
them incremental.
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
Stephen Bennett wrote:
> On Mon, 01 Jan 2007 17:24:03 -0800
> Donnie Berkholz <dberkholz@gentoo.org> wrote:
>
>> Stephen Bennett wrote:
>>> The proposal means that all variables listed in USE_EXPAND get
>>> handled exactly as USE does where profile inheritance is concerned.
>>> Subprofiles can add to and remove from the value in the parent
>>> profile just as they can for USE.
>> Did I misread what you said earlier?
>>
>> Stephen Bennett wrote:
>>> At present,
>>> from what zmedico told me, it's handled in a weird manner which
>>> essentially does half the job, letting you add flags but not remove
>>> them in subprofiles.
>
> "At present" -- that's the behaviour that I want to change by making
> them incremental.

Oh, I see. I thought you were describing the present behavior of the
change as it would affect USE_EXPAND. As long as I can do
VIDEO_CARDS="-vesa", I'm happy with it.

Thanks,
Donnie
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Mon, 1 Jan 2007 21:12:47 +0000
Stephen Bennett <spb@gentoo.org> wrote:

> Following a discussion in #gentoo-portage earlier this evening, it was
> suggested that I send out an RFC email for this. So, does anyone
> object to requiring that any variable listed in USE_EXPAND be treated
> as incremental, at least as far as profile inheritance is concerned?

The main problem I see is that if it's becoming a real incremental
(across all config layers) it would change behavior for users as they'd
need to prefix their use-expanded vars with -* to retain their current
config. But I guess that's why you added that final clause about
profile scope.

Marius
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Tue, 2 Jan 2007 21:14:19 +0100 Marius Mauch <genone@gentoo.org>
wrote:
| On Mon, 1 Jan 2007 21:12:47 +0000
| Stephen Bennett <spb@gentoo.org> wrote:
| > Following a discussion in #gentoo-portage earlier this evening, it
| > was suggested that I send out an RFC email for this. So, does anyone
| > object to requiring that any variable listed in USE_EXPAND be
| > treated as incremental, at least as far as profile inheritance is
| > concerned?
|
| The main problem I see is that if it's becoming a real incremental
| (across all config layers) it would change behavior for users as
| they'd need to prefix their use-expanded vars with -* to retain their
| current config. But I guess that's why you added that final clause
| about profile scope.

You know, making users do that wouldn't necessarily be a bad thing... I
know that Paludis users like having the option of inheriting defaults
from the profile for expanded vars...

--
Ciaran McCreesh
Mail : ciaranm at ciaranm.org
Web : http://ciaranm.org/
Paludis is faster : http://ciaranm.org/show_post.pl?post_id=61
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Tue, 2 Jan 2007 20:21:45 +0000
Ciaran McCreesh <ciaranm@ciaranm.org> wrote:

> On Tue, 2 Jan 2007 21:14:19 +0100 Marius Mauch <genone@gentoo.org>
> wrote:
> | On Mon, 1 Jan 2007 21:12:47 +0000
> | Stephen Bennett <spb@gentoo.org> wrote:
> | > Following a discussion in #gentoo-portage earlier this evening, it
> | > was suggested that I send out an RFC email for this. So, does
> anyone | > object to requiring that any variable listed in USE_EXPAND
> be | > treated as incremental, at least as far as profile inheritance
> is | > concerned?
> |
> | The main problem I see is that if it's becoming a real incremental
> | (across all config layers) it would change behavior for users as
> | they'd need to prefix their use-expanded vars with -* to retain
> their | current config. But I guess that's why you added that final
> clause | about profile scope.
>
> You know, making users do that wouldn't necessarily be a bad thing...
> I know that Paludis users like having the option of inheriting
> defaults from the profile for expanded vars...

Didn't say it's a bad idea, but it's one of those things that creates a
lot of noise.

Marius
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
Marius Mauch wrote:
> The main problem I see is that if it's becoming a real incremental
> (across all config layers) it would change behavior for users as they'd
> need to prefix their use-expanded vars with -* to retain their current
> config. But I guess that's why you added that final clause about
> profile scope.

I'd gotten the impression that users would have to do that anyway once
flameeyes gets his ALSA_CARDS expanded, what with all the cards that
were going to be on by default anyway. Weren't there some other
variables in make.conf that were going to behave this way at some point?
Though it'd be nice to have this done sanely; for example, configuring X
turns off most everything for the user already; no need for "-* nvidia
vesa" for example, it's a minor thing to add, long as users get
sufficient warning.

It's a pretty simple fix for the documentation, too, long as we know
what docs would be affected by all the expansion and incrementing goin' on.
--
gentoo-dev@gentoo.org mailing list
Re: RFC: making USE_EXPANDed variables incremental [ In reply to ]
On Tue, 2 Jan 2007 21:14:19 +0100
Marius Mauch <genone@gentoo.org> wrote:

> The main problem I see is that if it's becoming a real incremental
> (across all config layers) it would change behavior for users as
> they'd need to prefix their use-expanded vars with -* to retain their
> current config. But I guess that's why you added that final clause
> about profile scope.

Indeed. Personally, I quite like the idea of doing it that way, but I
can see how others might not want to change it. The motivation for
bringing this up came from writing PMS, which is why I'm only really
concerned about profiles -- that document isn't concerned about user
configuration, which is why I added the final clause.
--
gentoo-dev@gentoo.org mailing list