1 |
On Wed, 22 Nov 2006, Brian Harring wrote: |
2 |
|
3 |
> What is annoying there is that since package.keywords is under |
4 |
> *very* weird rules, you can't do the usual incremental tricks there. |
5 |
> |
6 |
> Try specifying that you want to run only stable arch via |
7 |
> package.keywords sometime; it's not possible. |
8 |
|
9 |
In any case, I'm not proposing any change to package.keywords or |
10 |
ACCEPT_KEYWORDS. Maybe I ought to, because the situation is confusing and |
11 |
some useful things aren't possible, but that's a separate issue. |
12 |
|
13 |
> You're proposing further overloading of ACCEPT_KEYWORDS; how would |
14 |
> this fly for package.keywords? |
15 |
|
16 |
I wouldn't change ACCEPT_KEYWORDS at all or anything in the computation of |
17 |
pgroups or mygroups in portdbapi.gvisible(), so package.keywords is |
18 |
unchanged and the whole incremental ACCEPT_KEYWORDS is also unchanged. |
19 |
|
20 |
The present documentation (Ebuild HOWTO, for example) says "If the |
21 |
KEYWORDS flag has a preceding -, then the package does not work with the |
22 |
given keyword." But this is ignored by portage, because "Packages that do |
23 |
not support the native architecture are automatically masked by Portage", |
24 |
which means that "-x86" has no effect, because an ebuild with "-x86" won't |
25 |
have "x86" or "~x86", and is therefore masked. I'm proposing that "-foo" |
26 |
be taken seriously as meaning that the package does not work with the |
27 |
"foo" keyword, even if the ebuild would otherwise be visible. |
28 |
|
29 |
> Alternative that comes to mind (and is cleaner) is to actually have |
30 |
> portage *know* it's doing cross compilation; eliminates a fair bit of |
31 |
> the nasty ROOT detection it has to do; that cleans up the innards a |
32 |
> bit, but for it to work you would need a RESTRICT=cross-compile to |
33 |
> mark packages that can't be x-compiled as unusable (yet another |
34 |
> visibility filter, yay!) :) |
35 |
|
36 |
I don't think adding a RESTRICT option that's a visibility filter, unlike |
37 |
all of the existing ones (which affect what steps in building a package |
38 |
portage can take), would clarify anything. Having portage automatically |
39 |
enable whatever visibility filter there is for ebuilds that don't |
40 |
cross-compile would be nice, but that's a separate step. |
41 |
|
42 |
> Downside, such a restrict isn't fine grained per arch; pkg may |
43 |
> cross-compile fine for arches running linux, but explode targettubg |
44 |
> gfbsd for example. |
45 |
|
46 |
That's an issue with any proposal, but I bet that actually getting that |
47 |
level of details into ebuilds is going to be impractical anyway. |
48 |
|
49 |
-Daniel |
50 |
*This .sig left intentionally blank* |
51 |
-- |
52 |
gentoo-portage-dev@g.o mailing list |