1 |
On Thu, 2006-05-11 at 19:39 +0200, George Hedfors wrote: |
2 |
> Well no, but I did this change before doing the actual bootstrap. I |
3 |
> followed the quick install guide and made changes to make.conf, then |
4 |
> completed bootstrap.sh and started emerging the system. Thats when it |
5 |
> happend. As there is a stage-installed perl version, when the library |
6 |
> tried to compile, it used the present perl version as it was dependent |
7 |
> on it. However, as the stage-installed version did'nt work with my |
8 |
> changes, i had to manually upgrade perl before i could emerge the rest |
9 |
> of the system. I'm sure other people will have the same problem and to |
10 |
> solve this, one way would be to reinstall perl somewhere in the very |
11 |
> beguinning of emerge system. |
12 |
> |
13 |
> I'm just a stupid user trying to make life easier for others :). |
14 |
|
15 |
While I understand and we truly do appreciate the input, this is |
16 |
*exactly* why Release Engineering has dropped support for stage1 and |
17 |
stage2 installs and recommend to everyone to use the stage3 tarballs for |
18 |
*all* installations. |
19 |
|
20 |
The stage3 tarball is a known-good minimal working system. |
21 |
|
22 |
There are simply too many variables possible in doing a stage1 or stage2 |
23 |
installation that need to be addressed. I honestly should change the |
24 |
end bootstrap.sh message to tell the user that if something breaks, they |
25 |
get to pick up the pieces themselves. ;] |
26 |
|
27 |
Gentoo has, unfortunately, become a victim of its own success. The more |
28 |
flexibility we add to the system, the harder it is to test for multiple |
29 |
scenarios that are possible when doing an installation. For some time |
30 |
now, it has been simply impossible for us to test stage1 and stage2 |
31 |
installations with any degree of certainty that what builds on my |
32 |
machine will build on yours. There's simply too many USE flags |
33 |
affecting the "system" target. Also, most people don't realize that |
34 |
"system" is 100% contrived by portage from the profile and its own |
35 |
internal dependency calculations. We simply can't make something appear |
36 |
sooner in the dependency tree than it already does, without editing the |
37 |
ebuilds, which almost always has unforseen consequences, causing a |
38 |
back-out. |
39 |
|
40 |
Trust me. This is exactly what takes us so long during releases to get |
41 |
going. It usually takes us a few weeks just to get the stages to build |
42 |
properly, due to changes that have happened in the tree since the last |
43 |
release. |
44 |
|
45 |
If you want an installation that actually *works*, use a stage3 tarball. |
46 |
If you're familiar with the concepts of bootstrapping and are |
47 |
comfortable with cleaning up after portage vomiting all over your system |
48 |
because of something that changed in the time since release and you |
49 |
doing your installation, feel free to use the lower stages. Personally, |
50 |
I *always* use a stage3 installation now. Most of the time, I use GRP. |
51 |
I hate waiting to get a usable system. I'd much rather be checking my |
52 |
email and letting a compile go on in the background than stare at a |
53 |
console of scrolling text and not be able to do much but ssh to another |
54 |
box or chat on IRC for days. |
55 |
|
56 |
-- |
57 |
Chris Gianelloni |
58 |
Release Engineering - Strategic Lead |
59 |
x86 Architecture Team |
60 |
Games - Developer |
61 |
Gentoo Linux |