Gentoo Archives: gentoo-dev

From: Tom Wijsman <TomWij@g.o>
To: gentoo-dev@l.g.o
Subject: Re: [gentoo-dev] Does the scm ebuild masking policy make sense for git?
Date: Mon, 15 Sep 2014 08:51:19
Message-Id: 20140915105048.00007b25@gentoo.org
In Reply to: Re: [gentoo-dev] Does the scm ebuild masking policy make sense for git? by hasufell
1 On Sat, 13 Sep 2014 22:44:49 +0000
2 hasufell <hasufell@g.o> wrote:
3
4 > Jauhien Piatlicki:
5 >
6 > > In the ideal country of elves. In the real life it can be not
7 > > possible to build and install software in a given distribution
8 > > without downstream patches. You can find examples of such live
9 > > ebuilds in Gentoo tree.
10 >
11 > I think it's not appropriate and shouldn't generally be done (with a
12 > few exceptions). If the live ebuild needs heavy patching to even
13 > work, then don't commit it to the tree.
14
15 Patches come and go, that has nothing to do with its presence in the
16 Portage tree. We shouldn't remove and add them based on the varying
17 amount of patches, as that is much work with an unfavorable result.
18
19 > > Anyway, summarizing, it is completely impossible to be sure that
20 > > live ebuild will be buildable for you on a given arch in the next
21 > > 15 min., even if it was so in the last 15 min.
22 >
23 > That goes for almost all ebuild variables. So you either drop the
24 > whole concept of live ebuilds or you do what is reasonable:
25
26 Given that some users specifically request them, the concept lives on.
27
28 > a) provide consistent ebuild information, including keywords
29
30 KEYWORDS="" is as consistent as it can get.
31
32 > b) ask upstream about their git workflow, which branches they use,
33 > what arches they even officially support
34
35 This is not how we fill in KEYWORDS.
36
37 > c) only add live ebuilds if the upstream git model is something that
38 > can be relied upon in one way or another and if you can keep up with
39 > the changes
40
41 The entire package presence depends on this, not just the live ebuilds.
42
43 > If your live ebuild breaks every 15 minutes, then it shouldn't be in
44 > the tree.
45
46 Bleeding edge is designed to break every 15 minutes; live ebuilds are
47 bleeding edge, some users do want bleeding edge in the Portage tree.
48
49 > I actually don't commit live ebuild unless I know that upstream is
50 > collaborative and I'v contributed to almost all of the projects which
51 > I package as live ebuilds.
52
53 Consider to step outside the ideal country of elves* and explore.
54
55 (* quoting Jauhien's words)