1 |
On Sunday 01 January 2006 02:39, Holly Bostick wrote: |
2 |
> And iirc, the virtual is not going to be linked to a hard-masked package |
3 |
> (or at least it most likely is not atm, or the hard-masked packages are |
4 |
> listed after the stable packages). |
5 |
> |
6 |
> So what I would do is check /var/cache/edb/virtuals and see what the |
7 |
> listing for jdk actually is. |
8 |
|
9 |
Ok, you're right off the track here. /var/cache/edb/virtuals hasn't been used |
10 |
(at all) since 2.0.50. Packages that provide virtuals are looked up when |
11 |
emerge is run and the internal list of those is essentially arbitrary. |
12 |
Virtuals listed in /etc/make.profile (and parents) are put after those in |
13 |
correct order. Virtuals listed in /etc/portage/profile/virtuals are put at |
14 |
the front. |
15 |
|
16 |
When there is a virtual dependency, these are checked in order for the first |
17 |
unmasked package that is able to satisfy the dependency. The fact that |
18 |
sun-jdk is being chosen means that it is the first package to have any |
19 |
unmasked versions available from the internal list. sun-jdk-1.4* being chosen |
20 |
means that sun-jdk-1.5* is still masked. |
21 |
|
22 |
sun-jdk-1.5* is in package.mask so adding dev-java/sun-jdk (or even |
23 |
=dev-java/sun-jdk-1.5*) to /etc/portage/package.unmask should fix that issue. |
24 |
It also has a KEYWORDS of "~x86 ~amd64" so adding the same |
25 |
to /etc/portage/package.keywords should fix that issue. Beyond that there |
26 |
shouldn't be a problem. |
27 |
|
28 |
-- |
29 |
Jason Stubbs |
30 |
-- |
31 |
gentoo-user@g.o mailing list |