| 1 |
I propose a new QA subproject, the TreeCleaners. |
| 2 |
|
| 3 |
This is a delicate subject for some developers, other developers don't |
| 4 |
care, and yet others want the cruft in the tree removed. The Tree |
| 5 |
Cleaning project's main goal is to identify broken and unmaintained |
| 6 |
packages in the tree and either get them fixed or mask and remove them. |
| 7 |
|
| 8 |
Criteria: |
| 9 |
1. Packages slated for removal must have no active maintainer. This is |
| 10 |
accomplished by looking in the package's metadata.xml for the maintainer |
| 11 |
tag. The maintainer tag must contain an active (non-retired) developer |
| 12 |
or team. The tree cleaners will maintain a list of ebuilds assigned to |
| 13 |
maintainer-needed; this list may end up on the web similar to Debian's |
| 14 |
WNPP[1]. A package with missing metadata.xml is assumed to be unmaintained. |
| 15 |
|
| 16 |
2. Packages slated for removal must have open bugs filled against them. |
| 17 |
It is not the policy of the QA team nor this subproject to remove |
| 18 |
packages because they have no maintainer. There are plenty of |
| 19 |
completely working packages in the tree with no maintainer; we are not |
| 20 |
trying to remove those. |
| 21 |
|
| 22 |
3. Packages slated for removal with simple to fix bugs may be fixed by |
| 23 |
the tree cleaners if a project member elects to do so. Many of the bugs |
| 24 |
are relatively minor ( depend fixes, revbumps, etc ) and could be done |
| 25 |
by someone given a bit of time. This isn't meant as a means to |
| 26 |
perpetually keep crap in the tree, moreso that in some cases minor bugs |
| 27 |
against a package are not grounds for removal. |
| 28 |
|
| 29 |
4. Preferably packages slated for removal shall have a dead or |
| 30 |
unresponsive upstream. An upstream that isn't interested in maintenance |
| 31 |
means more work for Gentoo in keeping the package up to date. For |
| 32 |
packages that already lack a maintainer in Gentoo, a dead upstream means |
| 33 |
there is no developer and no upstream for a package; aka no one to do |
| 34 |
the work. A dead upstream is not *required* however, crap ebuilds for |
| 35 |
packages with an active upstream are still valid to be removed if there |
| 36 |
are major bugs filed against them. |
| 37 |
|
| 38 |
5. Packages slated for removal shall have a last rites e-mail sent to |
| 39 |
the gentoo-dev mailing list. There will be no packages disappearing |
| 40 |
randomly out of the tree due to the tree cleaner project members. |
| 41 |
Transparency is key here, both on bugs, in package.mask, and on the |
| 42 |
mailing list. developers and users both need to know what is going on. |
| 43 |
|
| 44 |
6. Packages slated for removal shall have a 30 day period in |
| 45 |
package.mask prior to removal. This is tree cleaner policy, and it's |
| 46 |
one that I hope other developers will adopt. I've seen things pmasked |
| 47 |
and removed after a week, a "couple of days", or just pmasked and never |
| 48 |
removed. The 30 day period allows everyone using the package to see the |
| 49 |
masking message and the corresponding bug when they use portage. |
| 50 |
|
| 51 |
Questions and Comments are welcome, as always. |
| 52 |
|
| 53 |
-Alec Warner |
| 54 |
|
| 55 |
[1] http://www.debian.org/devel/wnpp/ |
| 56 |
-- |
| 57 |
gentoo-qa@g.o mailing list |
Archives