Gentoo Archives: gentoo-dev

From: Jeremy Olexa <darkside@g.o>
To: gentoo-dev@l.g.o
Subject: Re: [gentoo-dev] slot deps in package.mask and profiles
Date: Tue, 27 Jan 2009 05:18:34
Message-Id: 497E98F3.2060406@gentoo.org
In Reply to: Re: [gentoo-dev] slot deps in package.mask and profiles by Alec Warner
1 Alec Warner wrote:
2 > On Mon, Jan 26, 2009 at 8:01 PM, Jeremy Olexa <darkside@g.o> wrote:
3 >> Jorge Manuel B. S. Vicetto wrote:
4 >>> -----BEGIN PGP SIGNED MESSAGE-----
5 >>> Hash: SHA1
6 >>>
7 >>> Donnie Berkholz wrote:
8 >>>> On 21:04 Sun 25 Jan , Ciaran McCreesh wrote:
9 >>>>> On Sat, 24 Jan 2009 20:25:44 -0100
10 >>>>> "Jorge Manuel B. S. Vicetto" <jmbsvicetto@g.o> wrote:
11 >>>>>> I talked to Zac <zmedico> earlier in #gentoo-portage about adding an
12 >>>>>> entry to package.mask for KDE-4.2.0 using slot deps. Thomas
13 >>>>>> <tanderson> and Patrick <bonsaikitten> raised the concern we might
14 >>>>>> need profile eapis and that PMS nailed p.mask to EAPI-0.
15 >>>>>> Zac confirmed that the first stable version to support slot deps in p.
16 >>>>>> mask was 2.1.3.16, that it was stabled in bug 197165 - 14 months ago
17 >>>>>> - - and that the first stages to include it were the 2008.0 stages.
18 >>>>>> Thus, can we finally give the ok to use slot deps in package.mask? Can
19 >>>>>> we also give the ok to use it everywhere in all 2008.0 and later
20 >>>>>> profiles/ ?
21 >>>>> The Council approved profile eapi files for use a while ago (can't
22 >>>>> remember when -- http://council.gentoo.org/ isn't being updated), and
23 >>>> Last month's meeting
24 >>>>
25 >>>>> they discussed timeframes for using newer EAPIs then too. Did you see
26 >>>>> that discussion?
27 >>>> "An EAPI=0 profile always needs to exist so that users with old portage
28 >>>> can upgrade. Otherwise they will sync and have no valid profile available so
29 >>>> cannot emerge a new version of portage.
30 >>>>
31 >>>> "Decision: Approved. Existing stable profiles must use EAPI=0. New or dev
32 >>>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
33 >>>> profiles can use higher EAPIs.
34 >>> Acoording to this we will never be able to use slot deps in package.mask
35 >>> as it's a global file. Given my first mail, can we agree to make EAPI-1
36 >>> the minimum EAPI for files under profiles/ ? Can we also create a rule
37 >>> on how / when to update the minimum EAPI in profiles/ ?
38 >> So, portage that is unaware of EAPI-1 will just happily ignore the atom and
39 >> move on..? In that case:
40 >>
41 >> Please no! It is hard enough for a base 2007.0 install to be upgraded due to
42 >> the "portage & bash blocker" (and other issues) - We need to wait much
43 >> longer for an EAPI bump in a non-new profile (if ever, as Brian Harring
44 >> suggests - I agree).
45 >>
46 >> I know this might seem as a hassle to you but there *are* other entities
47 >> that provide a base 2007.0 install. Who knows how every
48 >> group/entity/company/etc use Gentoo.. While I agree that it isn't
49 >> necessarily our problem, however, we shouldn't make it harder for them or
50 >> anyone that has a 2007 base install. (We still mirror the 2007.0 stages[1],
51 >> 2007.0 cds are available[2] for purchase, etc[3] etc[4]).
52 >
53 > Dude, even people like Ubuntu/Canonical don't support stuff that old
54 > (current LTS is April 2008).
55 >
56 > The tree is now; see the date? It's 2009, not 2007.
57 >
58 > One of the biggest problems Gentoo has is backwards compatibility and
59 > legacy stuff; it is the nightmare of every project and there has to be
60 > a point where you say 'tough.' So make a decision, announce it widely
61 > that on X date the tree will just break for users; write up a FAQ on
62 > how to upgrade past it, and then make the changes.
63
64 2008.0 was released on Jul 6 2008[1]. So, you think that after 6 months,
65 it is time to say "tough"? Sorry, I don't agree.
66
67 [1]: http://www.gentoo.org/proj/en/releng/release/2008.0/index.xml#doc_chap2
68
69 >
70 > Realize once again that the tree was not designed very well and it has
71 > issues on a number of levels and it can't all be engineered around;
72 > and for progress to be made you will *have to break existing stuff*.
73 >
74 >> IMO, it would be a dis-service to bump EAPI in a non-new profile for our
75 >> user-base. I don't see any Pro's besides "easier to type" =/ So, I think the
76 >> Council decision is appropriate.
77 >
78 > You seriously see no benefits to EAPI 1 or 2 in profiles? What about
79 > slot deps? use deps? these things have been core feature requests
80 > since 2003; surely you don't think they are useless to our users?
81
82 No, I didn't say that at all, *sigh*
83
84 >
85 >> -Jeremy
86 >>
87 >> [1]: http://distfiles.gentoo.org/releases/x86/2007.0/
88 >> [2]: http://www.linuxcd.org/view_distro.php?lst=&id_cate=20&id_distro=12
89 >> [3]: http://lylix.net/linux-vps-plans
90 >> [4]: http://www.linode.com/faq.cfm
91 >>