1 |
Disclaimer: |
2 |
|
3 |
I am not a Gentoo developer; I have never been one; It is unlikely |
4 |
that I will ever be one. All of these opinions are my own. |
5 |
|
6 |
|
7 |
On 29 Mar 2007, nelson_batalha@×××××××.com spake: |
8 |
>> Catalyst builds are not supposed to be *interactive*. If you are |
9 |
>> hitting these problems, you're doing something wrong and/or |
10 |
>> stupid. Most of the time, getting blockers is a result of using an |
11 |
>> old stage3 seed to build a livecd with a newer snapshot. |
12 |
> |
13 |
> Actually some of them actually were bugs in the ebuilds, though most |
14 |
> errors I had were when I didn't knew of the package.use bug. Take |
15 |
> this one http://bugs.gentoo.org/show_bug.cgi?id=172622 |
16 |
|
17 |
Bugs in ebuilds are why the release engineering team is very careful |
18 |
about what snapshots they use when building releases. |
19 |
|
20 |
> Even if they are not supposed to be interactive, is there a good |
21 |
> reason for not allowing that? I think large emerges often results in |
22 |
> errors, no matter what stage3 seed is used. Besides, many emerges |
23 |
> are so much easier with interaction right? Take |
24 |
> http://forums.gentoo.org/viewtopic-t-539520.html |
25 |
|
26 |
In order for a release of any system to be "high quality", it's build |
27 |
process must first be "repeatable". That is, you should be able to |
28 |
take the same set of inputs and produce the same result -- this is a |
29 |
trivial requirement of the scientific method. Interactivity does not |
30 |
lend itself to repeatability. That is, things usually work better when a |
31 |
human is not one of the inputs. ;) |
32 |
|
33 |
If you are finding yourself needing interactivity in Catalyst, then |
34 |
you should step back and examine your process. If you've had to "help" |
35 |
catalyst with more than just configuration directives, then you likely |
36 |
don't have a "pure" result. |
37 |
|
38 |
I often have to re-try a build several times because of whacky Portage |
39 |
behavior or broken ebuilds, but because of pkgcache, ccache, and |
40 |
kerncache I rarely ever feel penalized for it. In fact, the most I |
41 |
usually have to do is *erase* selected items from the cache to force a |
42 |
rebuild. Only on very limited occasions have I had to mess with an |
43 |
"in-flight" build, and probably not at all since catalyst 2.x. |
44 |
|
45 |
Having been a professional swiss-army-geek my entire adult life, I've |
46 |
developed rather strong opinions about process control, software |
47 |
packaging, and system administration. I have to say that, for the most |
48 |
part, I think the Catalyst guys have the right idea. |
49 |
|
50 |
-- |
51 |
Stephen |