1 |
On 1/8/08, Ciaran McCreesh <ciaran.mccreesh@×××××××××××××.uk> wrote: |
2 |
> On Tue, 8 Jan 2008 18:44:22 -0800 |
3 |
> "Alec Warner" <antarus@g.o> wrote: |
4 |
> > > Uh... So where do the original problems come from? Are you saying |
5 |
> > > that packages mysteriously start breaking on their own because |
6 |
> > > no-one's maintaining them? |
7 |
> > |
8 |
> > Of course they do |
9 |
> |
10 |
> Ah, right. Because of the magical elf that lives in the CVS server |
11 |
> that mysteriously goes around breaking dependencies when no-one's |
12 |
> looking. |
13 |
> |
14 |
> Yes, a magical elf. Much more plausible than the theory that it's |
15 |
> actually developers screwing up by dropping keywords or best keyworded |
16 |
> version on a package's deps. |
17 |
|
18 |
I was going to go with 'the stable glibc changed' or 'some lib this |
19 |
software depended on was updated to a new version' or any other action |
20 |
that could cause software to not work as intended. |
21 |
|
22 |
I'm not trying to make the argument that developers don't screw up. |
23 |
Certainly mr_bones can attest that they do it on a daily basis. |
24 |
|
25 |
I think the argument here is that developers control ebuilds. If a |
26 |
given ebuild is causing 'trouble' for a maintainer it is within their |
27 |
control to remove the ebuild. Just as if a given package is causing |
28 |
the maintainer grief it can be deleted from the tree, so can keywords |
29 |
for a given arch be removed for a given ebuild (and possibly that |
30 |
ebuild removed because it is known to be old and buggy.) |
31 |
|
32 |
If the arch team wants that ebuild in the tree they should do some |
33 |
work to keep a given package up to date in terms of other arches or we |
34 |
should define some sort of metadata that notifies people that the arch |
35 |
team is the 'maintainer' for a given version of a package. |
36 |
|
37 |
I agree that you should not break the arch's tree by removing a given |
38 |
package (or it's last stable ebuild). |
39 |
-- |
40 |
gentoo-dev@l.g.o mailing list |