| 1 |
peter.gantner@×××××××××××××.at wrote: |
| 2 |
|
| 3 |
[snip] |
| 4 |
|
| 5 |
> In a LKML post [1] about the symlinking of /usr/include/linux to |
| 6 |
> /usr/src/linux/include/ which has been done for quite some time in older |
| 7 |
> Linux distributions, Linus calls doing this "instanity" and states that |
| 8 |
> the kernel headers used to compile glibc and friends should always be |
| 9 |
> taken from the glibc distribution (which includes a patched set of |
| 10 |
> kernel headers) and one should not use the headers from the kernel.org |
| 11 |
> tarball. |
| 12 |
> |
| 13 |
> But this is exactly what Gentoos linux-headers package does: get the |
| 14 |
> kernel tarball and install the required files. The result is in effect |
| 15 |
> not really different from the old symlinking method. |
| 16 |
|
| 17 |
On the contrary, it is quite different. As you suggested, the |
| 18 |
linux-headers package extracts 'include/asm/*' and 'include/linux/*' |
| 19 |
from the kernel tarball matching its release and places them under |
| 20 |
'/usr/include/'. Thus the headers are entirely independent from those in |
| 21 |
the sources you unpack to build your kernel and there are no symlinks to |
| 22 |
speak of. I'd say that's a fundamental difference. All of which is quite |
| 23 |
sensible and "correct", not least for the following reasons: |
| 24 |
|
| 25 |
1) There is no arbitrary requirement for a kernel source tree to reside |
| 26 |
at /usr/src/linux to build anything that's dependent on kernel headers |
| 27 |
(except for things like "out-of-tree" kernel modules such as |
| 28 |
nvidia-kernel which depend on the configuration of your particular |
| 29 |
kernel instance). |
| 30 |
|
| 31 |
2) The headers aren't going to "change" every time you change your |
| 32 |
preferred sources. |
| 33 |
|
| 34 |
3) Control of the headers versioning and release policy is placed in the |
| 35 |
hands of the vendor where it belongs and useful patches can be applied |
| 36 |
to fix known source build errors with quirky packages (as is the case |
| 37 |
with Gentoo's linux-headers package). |
| 38 |
|
| 39 |
Basically, the distro vendors have got it right and there is nothing to |
| 40 |
worry about, regardless of what an anachronistic post by Linus might |
| 41 |
lead one to believe ;) |
| 42 |
|
| 43 |
Regards, |
| 44 |
|
| 45 |
--Kerin Francis Millar |
| 46 |
-- |
| 47 |
gentoo-server@g.o mailing list |
Archives