1 |
Ciaran, |
2 |
|
3 |
It looks like a fairly trivial thing to fix in ebuilds. |
4 |
|
5 |
I think the problem you may be having is that people don't have any |
6 |
incentive to make short-term changes to their ebuilds just so you can |
7 |
get Paludis to work with them. It needs to be part of a larger |
8 |
interoperability plan that includes all interested parties, and serves |
9 |
the needs of everyone, not just the Paludis developers. |
10 |
|
11 |
-Daniel |
12 |
|
13 |
On 2/21/07, Ciaran McCreesh <ciaranm@×××××××.org> wrote: |
14 |
> A number of ebuilds do things like this: |
15 |
> |
16 |
> inherit foo |
17 |
> |
18 |
> VAR="${FOO_VAR}" |
19 |
> |
20 |
> src_compile() |
21 |
> { |
22 |
> emake blah=${VAR} |
23 |
> } |
24 |
> |
25 |
> where foo.eclass looks like this: |
26 |
> |
27 |
> foo_pkg_setup() |
28 |
> { |
29 |
> export FOO_VAR=baz |
30 |
> } |
31 |
> |
32 |
> and where VAR is usually one of the KV variables from linux-info. |
33 |
> |
34 |
> This works with Portage as it is currently implemented because Portage |
35 |
> re-sources everything between every phase. However, it breaks with pure |
36 |
> environment-saving based ebuild implementations, including Paludis and |
37 |
> possibly Portage in the future (pure environment saving is the only way |
38 |
> to avoid the removing eclass problems). So, the question: |
39 |
> |
40 |
> Is this something that has to work (at the expense of never allowing |
41 |
> eclass removal in the future), or is it a fluke and are ebuilds that |
42 |
> use it broken? An easy workaround is to move the ebuild global |
43 |
> variables into pkg_setup, and call the eclass_pkg_setup from within the |
44 |
> ebuild pkg_setup. |
45 |
> |
46 |
> -- |
47 |
> Ciaran McCreesh |
48 |
> Mail : ciaranm at ciaranm.org |
49 |
> Web : http://ciaranm.org/ |
50 |
> Paludis, the secure package manager : http://paludis.pioto.org/ |
51 |
> |
52 |
> |
53 |
> |
54 |
-- |
55 |
gentoo-dev@g.o mailing list |