1 |
Drake Donahue <donahue95@×××××××.net> posted |
2 |
19351.0605059148$1222999100@××××××××××.org, excerpted below, on Thu, 02 |
3 |
Oct 2008 21:56:58 -0400: |
4 |
|
5 |
> TRY using 2008.0 minimal livecd following the handbook in detail as |
6 |
> though you were a gentoo first timer. A lot of knowledge is also a |
7 |
> dangerous thing. |
8 |
|
9 |
Good advice! =:^) |
10 |
|
11 |
First, note that stage-1 installs are no longer supported, due to nasty |
12 |
circular dependency issues that are very difficult to work out. They're |
13 |
still provided and in theory can still work, but the decision was that it |
14 |
was simply too much trouble trying to support stage-1, given that the |
15 |
same customized end result can be achieved by starting from a stage-3, |
16 |
remerging system, then world. The caveat there is that one or more |
17 |
packages may need to be done out of order, if the USE flags changed |
18 |
significantly from the defaults for that stage-3. |
19 |
|
20 |
As for the glibc errors, what this type of glibc errors usually mean is |
21 |
that somewhere along the line your multilib config got screwed up, and |
22 |
gcc can no longer compile one of 32-bit or 64-bit correctly. There's a |
23 |
number of different ways the multilib could have gotten screwed, and it's |
24 |
often not worth the trouble to figure out which it was, but just to fix |
25 |
it, by starting from a stage-3 once again. Actually, it's often possible |
26 |
to fix the problem by remerging just one package, gcc, from the stage-3 |
27 |
(quickpkg it to a binpkg, then emerge -K the binpkg). However, just |
28 |
doing the full stage-3 should work as well and is more likely to fix |
29 |
other misc errors. |
30 |
|
31 |
FWIW, after losing 32-bit compiling several times, I got tired of it and |
32 |
went to the no-multilib profile, which then kills the 32-bit bits of |
33 |
gcc/glibc/binutils/sandbox. The biggest reason for multilib is legacy |
34 |
support of 32-bit-only closed source packages. Since they're closed |
35 |
source, porting them to 64-bit isn't an option (the major open source |
36 |
stuff has all long been ported, OpenOffice was one of the last open |
37 |
source apps not ported, and it is now), and 32-bit compatibility must be |
38 |
maintained if you use them. Since I don't do closed source (in general, |
39 |
I couldn't legally do so even if I wanted to, since I no longer agree to |
40 |
sign my rights away nor will I accept their disclaimer for damages when |
41 |
it's a black-box who /knows/ what's in), there's very little reason I'd |
42 |
ever need 32-bit, and avoiding all the hassles that come with multilib |
43 |
has been a MUCH better choice for me. |
44 |
|
45 |
Of course, even with no-multilib, if I needed to do 32-bit compiles as I |
46 |
now do since I just got an Acer Aspire One AOA-150L netbook and plan on |
47 |
putting Gentoo on it, with the work done on my main 64-bit no-multilib |
48 |
system, it's still very possible to install an x86 (32-bit) stage into a |
49 |
chroot and build a full 32-bit system from there. That's actually what |
50 |
I'll be doing, running a full 32-bit chroot with FEATURES=buildpkg, then |
51 |
installing the binary packages on my AA1 netbook. |
52 |
|
53 |
So if you can do without multilib, do consider switching to the no- |
54 |
multilb profile. It has certainly simplified my life here, and as a |
55 |
bonus, gcc and glibc only take half the time to merge, because they only |
56 |
build for a single bitness instead of two. Given that both those |
57 |
packages are fairly major and take a decent amount of time to build, |
58 |
halving that time is nice! |
59 |
|
60 |
-- |
61 |
Duncan - List replies preferred. No HTML msgs. |
62 |
"Every nonfree program has a lord, a master -- |
63 |
and if you use the program, he is your master." Richard Stallman |