kernel package building

ICMP Request icmp.request at gmail.com
Tue Jul 10 10:47:40 PDT 2007


Heya Ken!

> 
>  Why ?  Apart from when I'm in chroot, I always build in ~/ as a
> normal user.  If you run into kernel errors when building as root
> (seen that twice, I think) you will find that upstream doesn't
> really care.  For the book, the kernel version has been tested, so
> it shouldn't be a problem.
> 

/usr/src is only a convention, that's what I said "usually" at /usr/src 
. You can extract it wherever you want. I'm just following "standards".

>  Again, why ?  This used to be the way to build external modules,
> but shouldn't be required for 2.6.

Same answer: standards, convention, nothing necessary.

>  NO!  The headers in /usr/include are what glibc was compiled
> against.

/usr/include was always the place to put the kernel headers or am I 
wrong? Where do you put them?

>  If your use of 'raw' is the same as mine ("the kernel's own
> headers, not sanitised for userspace"), this is definitely a very
> bad idea for 2.6.  Fortunately, the kernel can now sanitise its own
> headers, which is what we do when we build LFS (fedora uses this,
> distros will undoubtedly move to this, at least for the main
> architectures).  But only the other day somebody had problems caused
> by repeating the book's instructions on a running system.

It's the same, as I was telling this method copies from 
kernel_source/include

I know that according to LFS-svn, the kernel headers are now OK for use. 
But particularly I don't trust on them yet. And, as I already told:

"This is just how I do it and it works fine for me. There are many other 
ways to compile the kernel."

You can invent your own way doing whatever you want, building the kernel 
from a directory other than /usr/src, putting the headers in a place 
other than /usr/include, etc. My way just follow standards, nothing 
more. The important is only one thing: to work fine. As long as it works 
fine and you don't get any stability errors or errors like header ones 
when compiling, it's fine and cannot be considered "wrong". (of course 
assuming that if you do in another way you wouldn't get this errors! ;] )

Regards,

-- 
ICMP Request - Don't forget to call me when you ping!

Mail/MSN: icmp.request at gmail.com

x86 Intel Prescott 32-bit Platform - Linux 2.6.20.1

Registered Linux User #449464



More information about the lfs-support mailing list