1 |
On 9/9/06, Mark Knecht <markknecht@×××××.com> wrote: |
2 |
> If it doesn't take too long, please, what is 'multilib'? I have a |
3 |
> full 32-bit chroot on my system since AMD64 doesn't support flash, |
4 |
> etc. Could this be related to that somehow? |
5 |
|
6 |
Maybe. Put simply, multilib is the feature of portage that rebuilds |
7 |
all (most?) libraries twice, once for /lib32, and once for /lib64, to |
8 |
support 32-bit and 64-bit applications respectively. |
9 |
|
10 |
Unless you are using a no-multilib sub-profile (which you are not), it |
11 |
is normal that sandbox will attempt to be built as a 32-bit version. |
12 |
The problematic point seems to be this test in the sandbox |
13 |
configuration: |
14 |
|
15 |
checking for i686-pc-linux-gnu-gcc... no |
16 |
|
17 |
If that says yes, for example because you have a cross-compiler |
18 |
installed, then you can have some problems to build the 32-bit version |
19 |
of sandbox. Thus why unmerging any cross-compilers or removing any |
20 |
stale i686-pc-linux-gnu-* links works around the problem. |
21 |
|
22 |
> lightning ~ # emerge -pv gcc |
23 |
> |
24 |
> These are the packages that would be merged, in order: |
25 |
> |
26 |
> Calculating dependencies... done! |
27 |
> [ebuild R ] sys-devel/gcc-4.1.1 USE="fortran gtk nls -bootstrap |
28 |
> -build -doc -gcj -hardened -ip28 -ip32r10k -mudflap -multislot -nocxx |
29 |
> -objc -objc++ -objc-gc -test -vanilla" 0 kB |
30 |
|
31 |
This bothers me slightly...my output has |
32 |
|
33 |
[ebuild R ] sys-devel/gcc-4.1.1-r1 USE="gcj (-altivec) -bootstrap |
34 |
-build -doc -fortran -gtk (-hardened) -ip28 -ip32r10k -mudflap |
35 |
(-multilib) -multislot (-n32) (-n64) -nls -nocxx -objc -objc++ |
36 |
-objc-gc -test -vanilla" 0 kB |
37 |
|
38 |
Notice the masked use flags multilib/n32/n64...you should be seeing |
39 |
the same things regardless of the profile. Maybe this is just due to |
40 |
differences in portage versions...I'm on 2.1.1 here. |
41 |
|
42 |
-Richard |
43 |
-- |
44 |
gentoo-user@g.o mailing list |