Index: 2.6-koutput-user.xml
===================================================================
RCS file: /home/cvsroot/gentoo/xml/htdocs/doc/en/2.6-koutput-user.xml,v
retrieving revision 1.1
diff -u -b -B -r1.1 2.6-koutput-user.xml
--- 2.6-koutput-user.xml 26 Mar 2004 00:25:16 -0000 1.1
+++ 2.6-koutput-user.xml 27 Mar 2004 00:33:41 -0000
@@ -21,8 +21,8 @@
+You may be asking yourself, "Why should I bother?" The problem that inspired
+this functionality was in fact an issue with kernel module ebuilds like
+nvidia-kernel or hostap-driver. These ebuilds need to use the kernel build
+system to compile, however building them would require updating files in
+
+Several dependant steps can be taken to resolve this issue. In short, kernel
+builds must output all of their files to a different location when they are
+compiled, and later external module compiles must also do this, only they
+output to some subdirectory of the temporary build location. By letting the
+modules output their built files into the permitted build location, those nasty
+sandbox errors are prevented.
+
-# config-kernel --output-dir=/home/joe/kernels +# config-kernel --output-dir /home/joe/kernels
@@ -128,7 +158,7 @@
-# config-kernel --output-dir=default +# config-kernel --output-dir default
@@ -138,13 +168,13 @@
-# config-kernel --make-koutput=/usr/src/linux-2.6.3 +# config-kernel --make-koutput /usr/src/linux-2.6.3
The above command converts the kernel found in
+This change of location is important. Two processes are largely affected by this
+new file location, copy a kernel config into the proper location when updating
+kernel sources, and copying a newly compiled bzImage into
+# cp /var/tmp/kernel-output/2.6.4/.config /var/tmp/kernel-output/2.6.5/ ++ +Copy our .config into the new output directory +# config-kernel --set-symlink 2.6.5 +# cd /usr/src/linux +Move into our new source tree +# make oldconfig +# make && make modules_install +Compile our new kernel and modules +# mount /boot +# cp /var/tmp/kernel-output/2.6.5/arch/i386/boot/bzImage /boot/ +# umount /boot +Install the new kernel image +
+For kernel modules building against kernels using the new system, nothing more +should be required beyond emerging the desired kernel module package as usual. +At the time of this writing, there are still some kernel modules being converted +over to the new setup, or still in testing phase. If the stable version of a +kernel module fails, try installing the ~arch version first: +
+ ++# ACCEPT_KEYWORDS='~x86' emerge foo ++ +The above command installs the testing version of the foo package +
+If this still fails, please open a bug at