1 |
Sven Köhler wrote: |
2 |
|
3 |
> Hi, |
4 |
> |
5 |
> the linux-headers installed the headers from a specific kernel version. |
6 |
> There seem to be a glibc-kernel-headers package out there. Search the |
7 |
> LKML for a mail with the subject "glibc-kernel-headers". |
8 |
> |
9 |
> I'm still very confused about the linux-headers ebuild. Linux Torvald |
10 |
> claims that a running system should not update the headers. OK, he only |
11 |
> recomms not to do that, but the portage offers be an update if i do |
12 |
> "emerge -upD world". I guess updating the headers could break some |
13 |
> things - although i cannot imagine which. |
14 |
|
15 |
Linus is correct in some aspects -- you shouldn't go blindly updating |
16 |
headers unless you're willing to risk breakage. |
17 |
|
18 |
Alot of distributions create headers by symlinking |
19 |
/usr/include/{linux,asm} to /usr/src/linux/include/{linux,asm}. Gentoo |
20 |
is one of the distributions that actually installs a separate set of |
21 |
headers in /usr/include, leaving you free to do whatever with |
22 |
/usr/src/linux (although it helps to keep this as a symlink to the |
23 |
source of your running kernel, some programs look for it). |
24 |
|
25 |
If you do upgrade your headers, all you really need to do is rebuild |
26 |
glibc and all should be well. building glibc against newer headers |
27 |
sometimes allows glibc to access new features of the kernel. The |
28 |
SELinux team recently needed to use 2.4.21 headers to access newer |
29 |
glibc/kernel features for SELinux because of this. Rebuilding the |
30 |
entire machine also doesn't hurt, but it can be overkill. |
31 |
|
32 |
The glibc-kernel-headers you speak of are "sanitized" headers mainly |
33 |
intended for redhat distributions and are prepared by Arjan van de Ven |
34 |
of Redhat. It was discussed several months ago on whether gentoo should |
35 |
use sanitized headers too, but our problem is we support multiple |
36 |
architectures, and I do believe the sanitized headers are for i386 only. |
37 |
sparc, mips, ppc, etc.. would all need certain things tweaked in their |
38 |
own headers as well for things to work properly. |
39 |
|
40 |
I think there is some thought on LKML to resolve this finally in 2.6, by |
41 |
creating a split in the kernel headers, separating out a sanitized set |
42 |
of user-space headers and headers intended exclusively for the kernel. |
43 |
No idea when this idea might be implemented in 2.6....it might become a |
44 |
2.7 feature. |
45 |
|
46 |
|
47 |
--Kumba |
48 |
|
49 |
-- |
50 |
"Such is oft the course of deeds that move the wheels of the world: |
51 |
small hands do them because they must, while the eyes of the great are |
52 |
elsewhere." --Elrond |
53 |
|
54 |
|
55 |
-- |
56 |
gentoo-dev@g.o mailing list |