1 |
> |
2 |
> On 14-05-2008 16:40:00 +0200, Markus Duft wrote: |
3 |
> > > I don't see how that's going to solve my problem of binutils tools (as |
4 |
> > > in the ebuild) being compiled such that if you want to run any of its |
5 |
> > > tools (e.g. objdump) it traps saying it can't find |
6 |
> > > libbfd.so.some.amazingly.long.version.number. |
7 |
> > |
8 |
> > I have really no idea, but it works here for every bootstrap. I |
9 |
> > believe (I think I mentioned this already) that the eprefix-bootstrap |
10 |
> > script sets the necessary runpaths (with CHOST queried from portageq). |
11 |
> > I think the relevant part is a function called need_prefix_env or |
12 |
> > so... |
13 |
> |
14 |
> That feels like an awfully ugly hack. Sorry to say so. Gives me an |
15 |
> idea on where/why/how though... |
16 |
|
17 |
I know it's an ugly hack, yes, and I feel it's really a bug in the binutils build mechanism, since that should set the rpath correctly for it's own prefix. But I'm not sure if there wasn't more than just binutils requiring the explicit runpath. Also this is just during bootstrap, so I can cope with that - in the real prefix the binutils-config does exactly the same thing, so.... |
18 |
|
19 |
Cheers, Markus |
20 |
|
21 |
> |
22 |
> |
23 |
> -- |
24 |
> Fabian Groffen |
25 |
> Gentoo on a different level |
26 |
> -- |
27 |
> gentoo-alt@l.g.o mailing list |
28 |
|
29 |
|
30 |
-- |
31 |
gentoo-alt@l.g.o mailing list |