Gentoo Archives: gentoo-amd64

From: Mark Knecht <markknecht@×××××.com>
To: gentoo-amd64@l.g.o
Subject: Re: [gentoo-amd64] Re: eselect problems
Date: Mon, 24 Nov 2008 22:15:38
Message-Id: 5bdc1c8b0811241415s74e385dbo25debbb6177bf9c7@mail.gmail.com
In Reply to: [gentoo-amd64] Re: eselect problems by Duncan <1i5t5.duncan@cox.net>
1 On Mon, Nov 24, 2008 at 11:59 AM, Duncan <1i5t5.duncan@×××.net> wrote:
2 > "Drake Donahue" <donahue95@×××××××.net> posted
3 > 414A2586925347C7BCFA0A784EC2A8FE@iwillxp333, excerpted below, on Mon, 24
4 > Nov 2008 12:19:04 -0500:
5 >
6 >> The wisdom of making currently existing and useful packages depend on
7 >> some future incomplete package management system (so that updates become
8 >> arduous and/or impossible)?? Anyone discovered a way to cope with
9 >> 'masked by eapx '? sys-apps/portage-2.2_rc15 did not relieve a 'masked
10 >> by eap2' ....
11 >
12 > It should be 2.2_rc16, now. rc15 has a circular logic bug on unsolved
13 > dependencies.
14 >
15 > But unless you happened to hit that bug, EAPI-2 should have worked, at
16 > least for anything in-tree and at least ~arch unmasked. I'm not sure why
17 > it wouldn't have and if it didn't it's certainly a bug, either in the
18 > package or in portage. Of course, overlays are an entirely different
19 > matter. In part, they are /intended/ to allow experimentation, and for
20 > the more experimental overlays, all bets and all guarantees are off. See
21 > below.
22 >
23 > But to answer your question, the EAPI restrictions work in stages, as
24 > follows:
25 >
26 > (1) Overlays can use whatever weird features they want, including ones
27 > only (for example) paludis supports, not portage at all. That's one of
28 > the reasons they are there, to allow the testing of experimental features.
29 >
30 > (2) In ordered to be unmasked to ~arch in the tree, the features used
31 > must be in a Gentoo council approved EAPI. The council has decided that
32 > it will only approve an EAPI (E-API, e being the traditional Gentoo
33 > prefix for package manager stuff and API having the traditional meaning)
34 > once it is concretely defined such that all three package managers agree
35 > on the definition, and when the official Gentoo PM, portage, supports it,
36 > at the same ~arch level as the packages depending on that EAPI.
37 >
38 > (3) No package using a new EAPI can go stable until the official package
39 > manager, portage, supports it in a stable version as well.
40 >
41 > If you are using overlay ebuilds, it's assumed you know how experimental
42 > that overlay is allowed to be and that you are willing to run an equally
43 > experimental package manager implementation of whatever features it
44 > uses. Gentoo neither can nor does attempt to make guarantees at that
45 > level. If you are using ~arch packages, there are limited guarantees
46 > attempted, but as the express purpose of ~arch is to allow packages
47 > intended for stable to be tested, but it's understood to be unstable and
48 > thus there will be bugs from time to time. Again, if you choose to use
49 > such packages, it's assumed you have your reasons and can manage to live
50 > with the consequences of the bugs that will occur.
51 >
52 > If you don't like those terms, only use stable, but be prepared to live
53 > with packages somewhat behind the times. In some cases, particularly
54 > with new hardware or with software that just added a very popular new
55 > feature, this will mean you'll simply do without support if you require
56 > that hardware support or feature to use the package, until the package
57 > works its way thru the system and is stabilized.
58 >
59 > Personally, I default to ~arch, and unmask hard-masked packages either in-
60 > tree or from various overlays from time to time as well. But in general,
61 > I'm prepared for them to fail once in awhile, and to spend sometimes
62 > significant time tracing and reporting bugs, then working around them or
63 > rolling back to an earlier version without the bug, should I need the
64 > bugged functionality. But YMMV definitely applies in this area, and
65 > while I'd not be satisfied running what I'd consider long outdated
66 > versions that may be the latest stable in many cases, other people may
67 > prefer that stability even if it does mean being maybe a year behind at
68 > times, possibly even more in some cases.
69 >
70 > --
71 > Duncan - List replies preferred. No HTML msgs.
72 > "Every nonfree program has a lord, a master --
73 > and if you use the program, he is your master." Richard Stallman
74 >
75 >
76 >