1 |
On Thu, 2008-11-27 at 10:14 +0000, Alan Hourihane wrote: |
2 |
> On Thu, 2008-11-27 at 11:01 +0100, Michael Haubenwallner wrote: |
3 |
> > On Wed, 2008-11-26 at 16:37 +0000, Alan Hourihane wrote: |
4 |
> > <snip> |
5 |
> > > > > Just static. |
6 |
> > > > |
7 |
> > > > Ohw, no shared libs! |
8 |
> > > > Wouldn't this require some (portage-)mechanism to perform a |
9 |
> > > > reverse-dependency-rebuild based on {,R}DEPEND? |
10 |
> > > > |
11 |
> > > > As far as I understand, get_libname() returns the extension used for |
12 |
> > > > *shared* libraries (which on AIX unfortunately can be '.a'). |
13 |
> > > > |
14 |
> > > > What should $(get_libname) return when there are no shared libraries? |
15 |
> > > > |
16 |
> > > > The test in readline just prevents moving 'lib*.a' out of /lib when it |
17 |
> > > > (potentially) is the shared library. |
18 |
> > > > |
19 |
> > > > When there are no shared libs at all, isn't it just wrong to have |
20 |
> > > > get_libname() returning '.a'? Maybe it should return either the empty |
21 |
> > > > string, or something like '.irrelevant' IMO. |
22 |
> > > |
23 |
> > > Having get_libname() return .a on FreeMiNT makes things behave correctly |
24 |
> > > for the majority of ebuilds. |
25 |
> > |
26 |
> > What would really break when get_libname() does not return '.a'? |
27 |
> > |
28 |
> > IIRC, some ebuilds have a 'static' useflag. |
29 |
> > What if this is use.forced in the FreeMiNT profiles? |
30 |
> |
31 |
> The -static USE flag in some of the ebuilds places some unnecessary |
32 |
> restrictions on FreeMiNT, which I'd probably end up having to workaround |
33 |
> anyway. |
34 |
> |
35 |
> Plus, readline doesn't have a static flag. So it's a moot point. |
36 |
> |
37 |
> > > Given AIX uses .a for shared libs makes me think that readline should |
38 |
> > > really be AIX specific and use CHOST == *-aix* rather than just globally |
39 |
> > > using != .a. |
40 |
> > |
41 |
> > /me just doesn't feel (yet?) get_libname() returning '.a' is really |
42 |
> > correct. |
43 |
> |
44 |
> I feel it is currently for FreeMiNT. |
45 |
> |
46 |
> If we can't make that change to readline (and libiconv), then I'll need |
47 |
> to revert back to suggesting... |
48 |
> |
49 |
> [[ $CHOST == *-mint* ]] && |
50 |
> |
51 |
> with exactly the same "mv" command as I can't have static libs in /lib. |
52 |
|
53 |
Let me try the ".irrelevant" flag and see what breaks though. |
54 |
|
55 |
Alan. |