1 |
Thanks Benda. At least there is a rational reason for what seemed to be a |
2 |
"hijacking". In my original post I noted that |
3 |
|
4 |
eselect profile list |
5 |
|
6 |
returned |
7 |
|
8 |
!!! Warning: Failed to determine ${ARCH}. Please submit a bug report. |
9 |
!!! Error: Cannot determine architecture |
10 |
!!! Error: Failed to get a list of valid profiles |
11 |
|
12 |
I was able to make changes to |
13 |
$EPREFIX/usr/share/eselect/libs/package-manager.bash |
14 |
so that I could see the standalone profile which now looks a lot better than the |
15 |
prefix-rpath profile. So now to emerging @system. With the new profile where |
16 |
will libs be installed - under $EPREFIX/usr/lib or $EPREFIX/usr/lib64? This |
17 |
presents a problem in installing coreutils with USE=acl. The sys-apps/acl libs |
18 |
get installed under $EPREFIX/usr/lib64 but apparently coreutils looks for them |
19 |
under $EPREFIX/usr/lib. And of courses libs were installed under |
20 |
$EPREFIX/usr/lib on my old rpath-prefix with glibc. My old linker may be at |
21 |
issue! |
22 |
|
23 |
Steven |
24 |
|
25 |
--- |
26 |
On Thu, 18 Feb 2016 13:22:57 +0900 |
27 |
Benda Xu <heroxbd@g.o> wrote: |
28 |
|
29 |
> Hi Steven, |
30 |
> |
31 |
> I am really sorry for the confusion in glibc on Prefix. We are on the |
32 |
> way to make it work out-of-the-box, finally. |
33 |
> |
34 |
> Steven Trogdon <strogdon@×××××.edu> writes: |
35 |
> |
36 |
> > OK, I think I have found part of my problem. Here is a partial result |
37 |
> > from updating world |
38 |
> > |
39 |
> > [ebuild R ] dev-libs/libltdl-2.4.6 ABI_X86="(-64*)" |
40 |
> > [ebuild R ] virtual/service-manager-0 KERNEL="(-linux*)" |
41 |
> > [ebuild R ] dev-libs/mpfr-3.1.3_p4 ABI_X86="(-64*)" |
42 |
> > [ebuild R ] dev-libs/mpc-1.0.3 ABI_X86="(-64*)" |
43 |
> > [ebuild N #] dev-libs/libiconv-1.14-r1 USE="static-libs" |
44 |
> > [ebuild R ] virtual/libiconv-0-r2 ABI_X86="(-64*)" ELIBC="(-glibc*)" |
45 |
> > [ebuild N #] dev-libs/libintl-0.19.7 USE="threads -static-libs" |
46 |
> > [uninstall ] sys-libs/glibc-2.19-r1 |
47 |
> > [blocks b ] sys-libs/glibc ("sys-libs/glibc" is blocking |
48 |
> > dev-libs/libiconv-1.14-r1) |
49 |
> > [blocks b ] sys-libs/glibc ("sys-libs/glibc" is blocking |
50 |
> > dev-libs/libintl-0.19.7) |
51 |
> > [ebuild R ] virtual/libintl-0-r2 ABI_X86="(-64*)" ELIBC="(-glibc*)" |
52 |
> > [ebuild R ] sys-libs/zlib-1.2.8-r1 ABI_X86="(-64*)" |
53 |
> > [ebuild R ] dev-libs/expat-2.1.0-r4 ABI_X86="(-64*)" |
54 |
> > [ebuild R ] dev-libs/iniparser-3.1-r1 ABI_X86="(-64*)" |
55 |
> > |
56 |
> > I don't know about the ABI changes, but notice that portage wants to |
57 |
> > remove sys-libs/glibc. This will break my prefix since I'm not using |
58 |
> > the host glibc but a prefix installed one. Of course things are pretty |
59 |
> > much broken now since I can't perform a meaningful update. Some |
60 |
> > ebuilds have the 'elibc_glibc' useflag to prevent installing things |
61 |
> > that conflict with an installed glibc - gettext is an example. That |
62 |
> > flag just doesn't work. I'm unable to install gettext (which is |
63 |
> > already installed) because of a file collision with a header provided |
64 |
> > by glibc. I have tried everything I can think of to enable the |
65 |
> > 'elibc_glibc' flag without success. I can certainly fix the ebuild to |
66 |
> > make it work but there remains the issue with glibc and the need for |
67 |
> > libiconv and libintl which are not needed but portage insists on |
68 |
> > installing? This has not been a problem until recently. |
69 |
> |
70 |
> What profile are you using? From your previous email it seems wrong: |
71 |
> |
72 |
> > Any ideas as to why I now have no valid ARCH set? The |
73 |
> > etc/portage/make.profile symlink is in place |
74 |
> |
75 |
> > ls -al ~/etc/portage/make.profile |
76 |
> > lrwxrwxrwx 1 strogdon math 71 Dec 24 00:17 |
77 |
> > /storage/strogdon/gentoo-prefix/etc/portage/make.profile -> |
78 |
> > /storage/strogdon/gentoo-prefix/usr/portage/profiles/prefix/linux/amd64 |
79 |
> |
80 |
> > Any pointers as to where to look? |
81 |
> |
82 |
> prefix/linux/amd64 is for prefix-rpath, which replies on host libc. |
83 |
> What should be used is prefix/linux-standalone/amd64, see bug 521138[1]. |
84 |
> |
85 |
> On my system, |
86 |
> |
87 |
> $ eselect profile list |
88 |
> Available profile symlink targets: |
89 |
> [1] gentoo_prefix:prefix/linux/amd64 |
90 |
> [2] gentoo_prefix:prefix/linux-standalone/amd64 * |
91 |
> |
92 |
> Therefore |
93 |
> |
94 |
> $ eselect profile set 2 |
95 |
> |
96 |
> would do the profile switch. |
97 |
> |
98 |
> > Now it's possible that my prefix is messed up but until recently my |
99 |
> > non-host glibc has been coexisting nicely with prefix. |
100 |
> |
101 |
> There is still hope to fix the running system. After switching to the |
102 |
> currect profile, emerge @system again. I am not sure if the glibc in |
103 |
> prefix overlay would work[2]. If not, give the one in my overlay[3] a |
104 |
> try. |
105 |
> |
106 |
> I am available to help with any further problems. |
107 |
> |
108 |
> Yours, |
109 |
> Benda |
110 |
> |
111 |
> 1. https://bugs.gentoo.org/show_bug.cgi?id=521138 |
112 |
> 2. https://bugs.gentoo.org/show_bug.cgi?id=473728 |
113 |
> 3. https://gitweb.gentoo.org/dev/heroxbd.git/tree/sys-libs/glibc |
114 |
> |