1 |
On Fri, 31 Aug 2012 13:03:00 -0700 |
2 |
Zac Medico <zmedico@g.o> wrote: |
3 |
> For those who may not know, chromium-os currently uses a |
4 |
> hard-host-depends ebuild as a workaround for our lack of HDEPEND |
5 |
> support [1]. We could easily add HDEPEND in EAPI 5 if we want, since |
6 |
> we already have a Portage patch attached to bug #317337 [2]. Here is |
7 |
> a summary of what that Portage patch will do: |
8 |
> |
9 |
> In EAPI 5 or later, DEPEND has been divided into two parts: |
10 |
> DEPEND for build-time target dependencies, and HDEPEND for |
11 |
> build-time host dependencies. This division is designed |
12 |
> specifically to minimize difficulty in the process of |
13 |
> adapting ebuilds that were written for earlier EAPIs, |
14 |
> and therefore it also minimizes the adjustments that |
15 |
> ebuild developers will have to make to the thought |
16 |
> processes involved when writing ebuilds from scratch. In |
17 |
> an environment that does not involve cross-compilation, |
18 |
> HDEPEND behaves the same as DEPEND. When an ebuild is |
19 |
> converted from EAPI 4 or earlier to EAPI 5 or later, |
20 |
> in order to support cross-compilation environments, some |
21 |
> dependencies may need to be migrated to HDEPEND. |
22 |
> |
23 |
> For ebuilds that have EAPI 5 or later, the emerge |
24 |
> --root-deps option has no effect since it is made obsolete |
25 |
> by division between DEPEND and HDEPEND. If EAPI 4 or |
26 |
> earlier ebuilds are used in combination with EAPI 5 or |
27 |
> later ebuilds, the --root-deps behavior will still be |
28 |
> applied to the EAPI 4 or earlier ebuilds (there is no |
29 |
> behavior change for ebuilds having older EAPIs). |
30 |
|
31 |
What exactly would the rules be for handling a package that is in both |
32 |
DEPEND and HDEPEND, when ROOT is in effect? Would the versions be |
33 |
expected to match? What about use flags? |
34 |
|
35 |
Also, we're getting rather a lot of *DEPEND variables here... If we're |
36 |
making people make major changes to their deps, which for HDEPEND we |
37 |
definitely would be, then the "it's expensive since people would have |
38 |
to redo their deps" argument against a combined DEPENDENCIES variable |
39 |
goes out of the window, so we should rethink that too. |
40 |
|
41 |
-- |
42 |
Ciaran McCreesh |