1 |
On Mon, Apr 24, 2017 at 04:07:03PM +0000, J. Roeleveld wrote: |
2 |
|
3 |
> >> I've had the odd rebuild failure here & there, don't bother re-emerging |
4 |
> >> it until revdep-rebuild has finished. Any dependencies it needs would |
5 |
> >> have been rebuilt by then and it should complete without problem. |
6 |
|
7 |
> Usually an emerge -e @world works. But this time the dependency tree has |
8 |
> some issues as I have had several failures. |
9 |
|
10 |
Here too. The revdep-rebuild only knows portage dependencies. But since it's |
11 |
a rebuild of existing packages, I think it doesn't see any reason to alter |
12 |
the order of merges. Thus we have the occasional failure because a |
13 |
linked-with lib has not yet been rebuilt, so you get an undefined reference. |
14 |
|
15 |
> Now just running the revdeprebuild line till it completes without error. |
16 |
> Then will try an emerge -e again. |
17 |
|
18 |
As I just found out: it will rebuild the entire list of packages again, |
19 |
because gcc-5.4.0 supplies the same libstdc++.so.6 as did gcc-4 (only with |
20 |
different ABI). To counteract that, I wrote the list of packages into a file |
21 |
and feed that to emerge in smaller chunks. That way I can keep tabs ob which |
22 |
packages have already been built (using genlop) and then remove those |
23 |
packages from my list file for subsequent emerge runs. |
24 |
|
25 |
-- |
26 |
Gruß | Greetings | Qapla’ |
27 |
Please do not share anything from, with or about me with any social network. |
28 |
|
29 |
Laziness is the habit of resting before you are tired. |