1 |
On Fri, 2008-11-28 at 14:08 +0000, Alan Hourihane wrote: |
2 |
> On Fri, 2008-11-28 at 11:35 +0100, Michael Haubenwallner wrote: |
3 |
> > On Fri, 2008-11-28 at 09:13 +0000, Alan Hourihane wrote: |
4 |
> > |
5 |
> > > > Eventually, sys-devel/libperl should generally be package.masked for |
6 |
> > > > FreeMiNT (as 'irrelevant', lol), or at least install nothing. |
7 |
> > > |
8 |
> > > I did try that a while ago, but it threw up lots of horrible problems |
9 |
> > > with DEPEND/RDEPEND. So what's the correct way to package.masked this to |
10 |
> > > install nothing ?? |
11 |
> > |
12 |
> > You could either package.provide it, or in the ebuild just 'return 0' in |
13 |
> > src_unpack(), src_compile() and src_install(). |
14 |
> |
15 |
> O.k. So maybe not so nice is that package.provide requires versions, so |
16 |
> I'd have to do.... |
17 |
> |
18 |
> sys-devel/libperl-5.8.8-r2 |
19 |
> |
20 |
> and keep updating that if a version bump happened, otherwise we'd end up |
21 |
> trying to emerge it and getting libperl.irrelevant again. |
22 |
|
23 |
So, I guess if the grobibot can change sys-devel/libperl to do this... |
24 |
|
25 |
[[ ${CHOST} == *-mint* ]] && return |
26 |
|
27 |
in each of the src_unpack, src_compile, src_install and pkg_postinst |
28 |
stages we're good to go here. |
29 |
|
30 |
And then the change to multilib.eclass can be done ".a" -> ".irrelevant" |
31 |
|
32 |
Alan. |