1 |
On Wed, 2007-11-21 at 10:09 +0100, Fabian Groffen wrote: |
2 |
> On 21-11-2007 08:49:12 +0100, Michael Haubenwallner wrote: |
3 |
> > On Tue, 2007-11-20 at 15:57 -0600, matt hull wrote: |
4 |
> > > i think we should use stable and testing in the current tree. no |
5 |
> > > different from other hurds. after 30 days with no major bugs and |
6 |
> > > tested mark the package as stable. as you said you dont have any |
7 |
> > > time, so you could just leave it unstable till there is time |
8 |
> > |
9 |
> > +1 |
10 |
> |
11 |
> Be aware that arch teams have a hard job doing stable markings, as the |
12 |
> entire dep chain needs to be stable too. In other words, I for sure |
13 |
> wouldn't put my time in it, so others would have to go this route. |
14 |
|
15 |
Agreed+accepted, eventually -0.5 then ;) |
16 |
|
17 |
> |
18 |
> > > i would like to see this for system packages so a change doesnt break |
19 |
> > > my system and force a reinstall. |
20 |
> > |
21 |
> > +2 |
22 |
> > |
23 |
> > Having at least system+svn stable would be really useful, especially for |
24 |
> > bootstrapping. |
25 |
> |
26 |
> I thought about this too, but what does it really add? The bootstrap |
27 |
> snapshot is already a freezed snapshot that is supposed to work (that's |
28 |
> why I can't just spin off a new snapshot, it needs to be in a stable |
29 |
> state). More than often I have to release a new snapshot just because I |
30 |
> fixed a quite show-stopping bug in the bootstrap image. A stable |
31 |
> keyword doesn't help here, as we either have to immediately bump these |
32 |
> packages into stable here, or the stable version causes trouble once an |
33 |
> update is being done. The nice library moves that I had to do are a |
34 |
> great example of this. |
35 |
|
36 |
Currently, snapshot needs to contain working ~arch system+svn packages |
37 |
for _all_ platforms at once, because if one snapshot is stable for one |
38 |
platform, it can turn out to be broken for another one, requiring a new |
39 |
snapshot, which might break the before-stable platform because it gets |
40 |
the new unstable packages even for bootstrap. |
41 |
|
42 |
/haubi/ |
43 |
-- |
44 |
Michael Haubenwallner |
45 |
Gentoo on a different level |
46 |
|
47 |
-- |
48 |
gentoo-alt@g.o mailing list |