| 1 |
-----BEGIN PGP SIGNED MESSAGE----- |
| 2 |
Hash: SHA1 |
| 3 |
|
| 4 |
Andrey Grozin wrote: |
| 5 |
> I'm using portage-2.2_pre*. After upgrading to sys-libs/readline-6.0, |
| 6 |
> portage (naturally) kept /lib/libreadline.so.5 -> |
| 7 |
> /lib/libreadline.so.5.2, because a lot of programs needed it. I did |
| 8 |
> emerge @preserved-rebuild, and only one binary using libreadline.so.5.2 |
| 9 |
> left: /usr/bin/gp, the interactive interpreter of sci-mathematics/pari. |
| 10 |
> Re-emerging it does not help: headers from readline-6 are used, but the |
| 11 |
> program links wirh libreadline.so.5.2. I thought something is wrong in |
| 12 |
> the pari's configure. This package does not use autoconf, but a |
| 13 |
> hand-written Configure script. It seems all right. Then I found that |
| 14 |
> /lib/libreadline.so symlink still points to libreadline.so.5. This |
| 15 |
> confuses the pari's Configure. I made it to point to libreadline.so.6 by |
| 16 |
> hand, and re-emerged pari. OK, gp now links with libreadline.so.6.0. Why |
| 17 |
> doesn't readline-6.0 ebuild install this symlink? |
| 18 |
> |
| 19 |
> After this re-emerging of pari, portage said |
| 20 |
> |
| 21 |
> !!! existing preserved libs: |
| 22 |
>>>> package: sys-libs/readline-6.0_p3 |
| 23 |
> * - /lib/libreadline.so |
| 24 |
> * used by /usr/bin/M2 (sci-mathematics/Macaulay2-1.2-r3) |
| 25 |
> * used by /usr/bin/Singular-3-0-4 (sci-mathematics/singular-3.0.4.4) |
| 26 |
> * used by /usr/bin/asy (media-gfx/asymptote-1.86) |
| 27 |
> * used by 111 other files |
| 28 |
> |
| 29 |
> I already re-emerged these packages (at least, 3 shown) as a part of |
| 30 |
> emerging @preserved-rebuild after upgrading readline, and they were not |
| 31 |
> in the @preserved-rebuild set before I changed the libreadline.so |
| 32 |
> symlink. Does this mean that all of these packages used libreadline.so |
| 33 |
> -> libreadline.so.5? This is not good. I'm going to emerge |
| 34 |
> @preserved-rebuild again, to be sure that my system is in a |
| 35 |
> self-consistent state. |
| 36 |
> |
| 37 |
> It seems that keeping libreadline.so -> libreadline.so.5 after merging |
| 38 |
> readline-6.0 and unmerging readline-5.2 is a bad idea. |
| 39 |
> |
| 40 |
> Andrey |
| 41 |
> |
| 42 |
> |
| 43 |
|
| 44 |
I'm not completely sure, but I think it doesn't install that symlink |
| 45 |
because it (correctly) installs a /usr/lib/libreadline.so ldscript, that |
| 46 |
tells the linker to link against /lib/libreadline.so.6, so long as the |
| 47 |
linker looks at /usr/lib before /lib, which is usually the case, unless |
| 48 |
"-L/lib" is passed to ld (by way of gcc). If you remove |
| 49 |
/lib/libreadine.so, then everything will Just Work, and portage will |
| 50 |
forget that that file ever existed as soon as you merge a package (any |
| 51 |
package). I'm not sure what the correct behavior of |
| 52 |
FEATURES="preserve-libs" should actually be in this case, but I think |
| 53 |
portage may need to look a little closer at the SONAME, and only |
| 54 |
preserve that file (like the old preserve_old_lib) or preserve only that |
| 55 |
symlink and what it links to, not everything that points at that. |
| 56 |
|
| 57 |
- -- |
| 58 |
Jonathan |
| 59 |
|
| 60 |
-----BEGIN PGP SIGNATURE----- |
| 61 |
Version: GnuPG v2.0.11 (GNU/Linux) |
| 62 |
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org |
| 63 |
|
| 64 |
iEYEARECAAYFAkq9rroACgkQOypDUo0oQOrUjgCdEoFum1tqM1f/g2BiGl7QKMc5 |
| 65 |
Ti0AoIeTDN6exLzTuR9hjRHe0UzKh41h |
| 66 |
=rwgK |
| 67 |
-----END PGP SIGNATURE----- |
Archives