| 1 |
On Sonntag, 24. April 2011, Samuli Suominen wrote: |
| 2 |
> On 04/24/2011 10:43 PM, Matthias Schwarzott wrote: |
| 3 |
> > Getting that discussion back on top. |
| 4 |
> > |
| 5 |
> >> Which is wrong, it should be /lib/foo instead, not $(get_libdir), to |
| 6 |
> >> follow what udev and other software in Linux has been using for a very |
| 7 |
> >> long time now. |
| 8 |
> > |
| 9 |
> > Sounds like we should fix udev ebuild and some ebuilds installing udev |
| 10 |
> > rules to not use /$(get_libdir)/udev, but plain /lib/udev. |
| 11 |
> |
| 12 |
> Right, doesn't make sense to have both 32bit and 64bit ELF's for udev, |
| 13 |
> so we should stick with /lib/udev. |
| 14 |
> |
| 15 |
> > I used that in believe that /lib is identical or links to /$(get_libdir) |
| 16 |
> > and multilib-strict requires it, but it seems to be intelligent enough |
| 17 |
> > to only deny 64-bit libs to go to /lib. |
| 18 |
> > |
| 19 |
> > So proper udev should use /lib/udev, correct? |
| 20 |
> |
| 21 |
> Correct. |
| 22 |
> |
| 23 |
> |
| 24 |
> |
| 25 |
> The udev situation is really a mess tree-wide, we have ebuilds |
| 26 |
> installing into 3 different directories now: |
| 27 |
> |
| 28 |
> /etc/udev (where user puts his local rules) |
| 29 |
> /$(get_libdir)/udev (as explained above) |
| 30 |
> /lib/udev (the correct one) |
| 31 |
> |
| 32 |
> Check the Portage to see the sad status of inconsistency: |
| 33 |
> |
| 34 |
> $ grep -r 'etc.*udev' */*/*.ebuild |
| 35 |
> $ grep -r 'get_libdir.*udev' */*/*.ebuild |
| 36 |
|
| 37 |
And this does not even catch the cases where Makefiles (eventuelly together |
| 38 |
with configure-parameters) install to any of these three locations. |
| 39 |
|
| 40 |
By the way, the bug that led me to think about the install location is this |
| 41 |
Bug #363549 |
| 42 |
|
| 43 |
Matthias |
Archives