1 |
>>>>> On Thu, 4 Jan 2018, Mike Gilbert wrote: |
2 |
|
3 |
> On Thu, Jan 4, 2018 at 5:23 AM, Pacho Ramos <pacho@g.o> wrote: |
4 |
>> I have seen this is only used by: |
5 |
>> app-arch/xz-utils |
6 |
>> dev-libs/gmp |
7 |
>> dev-libs/libpcre |
8 |
>> dev-libs/mpc |
9 |
>> dev-libs/mpfr |
10 |
>> net-nds/openldap |
11 |
>> sys-libs/gdbm |
12 |
>> sys-libs/ncurses |
13 |
>> sys-libs/readline |
14 |
>> sys-process/audit |
15 |
>> |
16 |
>> Maybe we could deprecate it and try to drop it in the future :/ |
17 |
|
18 |
> As Soap touched on earlier, this should probably not be |
19 |
> deprecated/removed until a solution compatible with Paludis and |
20 |
> pkgcore is implemented. |
21 |
|
22 |
> A couple of options for that: |
23 |
|
24 |
> 1. Add functionality similar to preserve-libs to these alternate |
25 |
> package managers. This is unlikely to happen. |
26 |
|
27 |
There may also be Portage users without preserve-libs in FEATURES. |
28 |
|
29 |
> 2. Slot the libraries so that the old versions may remain installed |
30 |
> in a PMS-compatible way. This is often a pain to actually implement. |
31 |
|
32 |
I don't think that this would fly. You'd have to split packages like |
33 |
xz-utils which install binaries, otherwise there would be collisions |
34 |
between slots. |
35 |
|
36 |
> If neither of these things happens, keeping the preserve_old_lib |
37 |
> calls in place is an ugly, but workable solution. |
38 |
|
39 |
+1 |
40 |
|
41 |
Also note that the eclass functions act as no-ops when they find |
42 |
preserve-libs in FEATURES. |
43 |
|
44 |
Ulrich |