1 |
-----BEGIN PGP SIGNED MESSAGE----- |
2 |
Hash: SHA512 |
3 |
|
4 |
Am Sonntag, 12. April 2015, 02:42:20 schrieb Alan Grimes: |
5 |
> |
6 |
> /etc/portage # emerge --info |
7 |
> [[: error while loading shared libraries: libtinfo.so.5: cannot open |
8 |
> shared object file: No such file or directory |
9 |
> Failed to validate a sane '/dev'. |
10 |
> bash process substitution doesn't work; this may be an indication of a |
11 |
> broken '/dev/fd'. |
12 |
> /etc/portage # |
13 |
|
14 |
OK. That means your system is seriously broken now, but I gather you already |
15 |
know that. I doubt there is a way to get back to a working system without |
16 |
tricks/extra downloads. |
17 |
|
18 |
Most of the comments below are now mostly "trying to find out what happened". |
19 |
I'm leaving out stuff where I dont see anything problematic. |
20 |
|
21 |
> > > The ncurses problem has been a low-level |
22 |
> > > issue for a long time but, with tinfo set, 99% of packages worked. |
23 |
> > |
24 |
> > Define low-level issue. What was broken all the time that you ignored? |
25 |
> |
26 |
> Busybox, valgrind, a number of other minor packages. |
27 |
|
28 |
OK, now what was wrong there? Do you still have any old build logs maybe? |
29 |
|
30 |
> > > 2. It sent out a profile that sets variable ABI_x86 with 32 bit |
31 |
> > > enabled. ALARM: ABI_x86 should be set in exactly one place: |
32 |
> > > /etc/portage/make.conf and nowhere else. But, nevertheless, ABI_x86 WAS |
33 |
> > > set which broke the profile because my system cannot compile 32 bit |
34 |
> > > executables. =( I tried the no-multilib profile but it didn't have a |
35 |
> > > number of essential useflags and was foobar. |
36 |
> > |
37 |
> > You're still not providing the slightest bit of useful information. |
38 |
> > |
39 |
> > What happens if you try to generate a 32bit executable? |
40 |
> |
41 |
> Fails completely due to linking errors even though it should always be |
42 |
> possible to compile something without its binary dependencies (with only |
43 |
> the headers) because symbol resolution should take place at load time. =| |
44 |
|
45 |
Well... Yes it is possible to compile things. But they also need to get |
46 |
linked, and for that the libraries need to be present. No you cannot build |
47 |
something against a library without the library being present. |
48 |
|
49 |
> CFLAGS="-O3 -march=native -pipe " |
50 |
> CXXFLAGS="${CFLAGS}" |
51 |
|
52 |
- -O3 is not a good idea. -O2 is safe. |
53 |
|
54 |
> LDFLAGS="-lncurses " |
55 |
|
56 |
I hope you forced that in only at the end when things were already broken and |
57 |
you tried to fix them. Wrong. Bad. Wrong. Bad. |
58 |
|
59 |
> ACCEPT_KEYWORDS="amd64 ~amd64" |
60 |
|
61 |
Should not be a big problem (since ~amd64 is pretty well maintained these |
62 |
days), but in general you'll likely hit more bugs if you run unstable. In this |
63 |
case it's definitely an advantage if you are familiar with bug reporting |
64 |
procedures and frequent our bugzilla. |
65 |
|
66 |
> USE="... |
67 |
> c++0x |
68 |
|
69 |
No. Just no, since you may run into all sorts of trouble with C++ programs. |
70 |
(There was a news item about this recently.) I doubt that it is related to |
71 |
your problems though. |
72 |
|
73 |
> custom-cflags |
74 |
|
75 |
No. Again. Just no. Do this and you get to keep the pieces. I dont dare to do |
76 |
that myself. Especially in combination with -O3 it's, well,... |
77 |
|
78 |
> discouraged |
79 |
|
80 |
Read it and despair. |
81 |
|
82 |
> sse2 sse3 sse4 |
83 |
|
84 |
Unproblematic but outdated. Also a recent news item. |
85 |
|
86 |
So, now about getting back to a working system. There are many ways, none of |
87 |
them really 100% clean given the amount of breakage, and I'm not really a |
88 |
specialist. Some ideas, unverified and untested; I've never done this myself. |
89 |
|
90 |
* Boot from a rescue system |
91 |
* Mount your gentoo installation somewhere |
92 |
* Copy your setup (/etc, /var/lib/portage/world) to a safe place (usb stick) |
93 |
* For an amd64 multilib system doublecheck that both in /usr and / the "lib" |
94 |
entry is a symlink to "lib64" in the same directory. |
95 |
* Download the newest amd64 multilib stage3 and untar it over your system. |
96 |
* restore your make.conf, *sanitized* |
97 |
CFLAGS="-march=native -O2 -pipe" |
98 |
* chroot into the system, and run emerge -eav system |
99 |
* then reboot normally and run emerge -eav world (you may have to repeat this |
100 |
a few times, since some packages may still be broken and portage won't |
101 |
necessarily rebuild them in the correct order) |
102 |
* then restore your old world file and repeat the last step |
103 |
|
104 |
IF YOU HAVE A SPARE PARTITION, there is a much better way- do a new install |
105 |
there, instead of all the above steps. |
106 |
|
107 |
Good luck. |
108 |
|
109 |
- -- |
110 |
Andreas K. Huettel |
111 |
Gentoo Linux developer (council, perl, libreoffice) |
112 |
dilfridge@g.o |
113 |
http://www.akhuettel.de/ |
114 |
-----BEGIN PGP SIGNATURE----- |
115 |
Version: GnuPG v2 |
116 |
|
117 |
iQIcBAEBCgAGBQJVKn2dAAoJEB9VdM6hupKV2ZsP/Asgl66mmh8/Lzd0mVwNQkts |
118 |
GSszMvt7RNUb2k4lQetXR0psyr9Up7NDrp0plmAAguNJMtZOdgIVZsJ5sTvtWDxT |
119 |
0fxM7EXD2QLzpKIUF0uQFLQ1O+68GYTK0UOai+7OzN99HwALM7U5H9k9hYIrEVY6 |
120 |
ZtueUazbfOgWmAM//ZPQltj4GYBeSiW9TLJUbZNHzQFqJkP+uPimYWN9QIkcGyVP |
121 |
UJUlU1IJP9IOF7Ing10zisK1CT3+tEXnyeU930bnaCW1rLAUcm797BmwKWZX9RfW |
122 |
MYqTe6jjWAmnK+n6ScAeCtcwmwZ5MeQQBD7SK8FySSpegev2dEZzrHaI9hvMajtZ |
123 |
4L3i/CCwSwxUs3hjh+cn9kCb1Cs6BUqtXfR9Jt/rSZFOWVE7TbRrIssW0ZEw92x4 |
124 |
jKAO//H41EjK830QB6WzWLJxfYSdQ1SEaEW2WHNSdYd4vaxHv9YQp3sdm3zKxWel |
125 |
hY0pEJ5NtQMWGkte0XU0DgowAuDnBMn7Xac//rcOwWsSZiWIn8SmdpxvR43/xAZt |
126 |
Ln4k7xxJhb3a4w9Uocm3vPMOxIgRo2xFrdYNN18VGGz9zVNNyOZhECJDy9EZEtF1 |
127 |
liswCUhLUQMhKBau8F/DSpoL4AzZzyHciMNVVVG5PbAeYCr749KKjyp5H7zmPwhC |
128 |
Vbe4Ykem3g/4BnMW1+6e |
129 |
=AjWC |
130 |
-----END PGP SIGNATURE----- |