1 |
yac posted on Thu, 13 Jun 2013 12:15:26 +0200 as excerpted: |
2 |
|
3 |
> On Wed, 12 Jun 2013 15:31:57 -0700 Greg Turner <gmt@×××××.us> wrote: |
4 |
> |
5 |
>> Anyhow, isn't the gentoo-x86 tree already plenty big enough, without |
6 |
>> every single overlay's ebuilds and eclasses in there too? Personally, |
7 |
>> I'm inclined to wish it was smaller, even if that meant more stuff was |
8 |
>> pushed into overlays |
9 |
> |
10 |
> Actually, this is something I expected to happen soon after we got |
11 |
> overlays but for some reason it haven't. I imagine we would not have a |
12 |
> single gx86 official tree but rather a bunch of official overlays. For |
13 |
> basic installation one would need just the system overlay. Then |
14 |
> everypony could add official overlay for KDE, or gnome or whatnot as one |
15 |
> desires. |
16 |
> |
17 |
> I haven't thought this through in any way but it feels like better |
18 |
> design. |
19 |
|
20 |
Someone else already mentioned the problem with that. At least |
21 |
currently, only the official tree is tested against, so at least in |
22 |
theory it's quite easy to have conflicts between overlays, and it's |
23 |
certainly much more likely to have packages broken for some usage as they |
24 |
simply haven't been tested against packages in that overlay. |
25 |
|
26 |
The more overlays, the more likely the conflicts and breakage. |
27 |
|
28 |
The obvious way around that is to have a set of "blessed" overlays that |
29 |
get tested against, much like the main tree (only) today. However, that |
30 |
seriously complexifies (good) testing as now every dev has to pull down |
31 |
the whole set of "blessed" overlays instead of just the main tree plus |
32 |
whatever overlays he happens to work on (with some devs doing no overlays |
33 |
at all), as is the case now. |
34 |
|
35 |
The last thing we should be doing is throwing additional roadblocks into |
36 |
the way of reasonable testing, and I believe that's why the split you |
37 |
expected hasn't happened -- people realize that and decide the main |
38 |
tree's the best idea after all. |
39 |
|
40 |
Tho CVS is a enough of a pain that I'm sure that alone keeps some |
41 |
packages and potential devs away. Once it's git, that problem too will |
42 |
disappear, and there will be less pressure to split off overlays than |
43 |
there is now. |
44 |
|
45 |
-- |
46 |
Duncan - List replies preferred. No HTML msgs. |
47 |
"Every nonfree program has a lord, a master -- |
48 |
and if you use the program, he is your master." Richard Stallman |