Gentoo Archives: gentoo-dev

From: Alain Penders <alain@g.o>
To: gentoo-dev@g.o
Subject: Re: [gentoo-dev] Stability and package mask
Date: Sun, 09 Mar 2003 19:54:45
Message-Id: 20030309195443.GF24326@purematrix.com
In Reply to: [gentoo-dev] Stability and package mask by Jim Nutt
1 On Sun, Mar 09, 2003 at 10:27:11AM -0800, Jim Nutt wrote:
2 > I noted in another message that the developers are looking for a way to
3 > get rid of package.mask in the 2.1 version of portage. I have an idea
4 > that might accomplish that, while still allowing the flexibility that
5 > Gentoo is prized for.
6 >
7 > Essentially, it would involve assigning each ebuild a numeric stability
8 > value. 10 would be the most stable, say equivalent to Debian stable, 5
9 > would be the equivalent of Debian testing and 0 (or lower) would be
10 > like Debian unstable. Each user would have a stability threshold set in
11 > their make.conf, packages with a lower stability number than the
12 > threshold would not be installed. This would allow each user to set
13 > their own comfort level.
14 >
15 > Every ebuild would start out at at 0 (really bleeding edge stuff
16 > might be negative!). Each week (or some other period of time) the
17 > ebuild is released without a bug reported, it's stability rating is
18 > increased by one point, until it reaches 10. If a bug is filed against a
19 > package, it remains at the same stability level until all the bugs filed
20 > against it are closed or it is manually bumped.
21 >
22 > Does this make sense to anyone else?
23
24 This only works for moving ebuilds from ~arch (unstable) to arch (stable)
25 masking. Several schemes similar to yours have been proposed in the past, and
26 so far I haven't seen any scheme that worked all the time. In your proposed
27 scheme, it would take 10 weeks for a package to become "stable", which is too
28 long for some, and too short for others. Add to that the threshold at which
29 a package is actually installed (by default, does it become mainstream when
30 it's 6 or 8 or...?), and you still create a critical point where all hell can
31 break loose... possibly several weeks after the creator released it and
32 forgot all about it.
33
34 package.mask is another issue entirely as it only contains packages that are
35 known to break system parts, and they won't be removed until those problems
36 are fixed. (E.g. Berkeley DB 4.* is in package.mask because installing it
37 breaks several other packages. Berkeley DB 4.x itself is stable.)
38
39 Alain
40
41 --
42 gentoo-dev@g.o mailing list