Bug 392769 – app-emulation/virtualbox-guest-additions-4.1.6-r1 fails to build from /etc/kernel/postinst.d - make[2]: *** No rule to make target `arch/x86/boot/Makefile.include.footer'. Stop.

• Status: RESOLVED OBSOLETE
• Product: Gentoo Linux

[Description Dennis Schridde 2011-12-02 01:05:46 UTC]

uname -r: 3.1.1-gentoo
/usr/src/linux -> linux-3.1.4-gentoo
Makefile:20: arch/x86/boot/Makefile.include.header: No such file or directory

Out of experience: This will succeed to emerge when I booted into 3.1.4.

build.log attached.

Reproducible: Always

[Comment 1 Dennis Schridde 2011-12-02 01:06:51 UTC]

Created attachment 294465 [details]
build.log (app-emulation/virtualbox-guest-additions-4.1.6-r1)

[Comment 2 Dennis Schridde 2011-12-02 21:14:09 UTC]

Kernel's "make install", executed in /usr/src/linux, builds this package via /etc/kernel/postinst.d/10_module-rebuild:
#!/bin/sh
exec emerge @module-rebuild

[Comment 3 Dennis Schridde 2011-12-02 21:21:51 UTC]

Actually it also happens when uname matches the /usr/src/linux sources. Sorry for the confusion. Apparently postinst.d is executed before those header files are created, which should probably be fixed.

[Comment 4 Jeroen Roovers (RETIRED) 2011-12-04 04:59:46 UTC]

Comment on attachment 294465 [details]
build.log (app-emulation/virtualbox-guest-additions-4.1.6-r1)

Er, if it's XZ compressed then you could simply set the MIME Type accordingly so bugzilla clients know what to expect instead of making humans work around the problem you have just created by helpfully pointing out that the MIME type is wrong. :)

[Comment 5 Dennis Schridde 2012-03-14 22:37:54 UTC]

Created attachment 305385 [details]
build.log (x11-drivers/ati-drivers-12.2)

x11-drivers/ati-drivers-12.2 appears to show the same (or at least a very similar) problem.

[Comment 6 Ben Kohler 2013-10-02 22:08:48 UTC]

Dennis, are you still having problems with this?  It does smell like a problem with inconsistent kernel sources or headers, but plenty of time has gone by that anything like that would gone away long ago.

We have a lot of aging vbox bugs and I'm trying to get an idea of which ones still need work, and which ones have faded away with time.

Thanks!

[Comment 7 Dennis Schridde 2013-10-03 11:07:37 UTC]

(In reply to Ben Kohler from comment #6)
> Dennis, are you still having problems with this?  It does smell like a
> problem with inconsistent kernel sources or headers, but plenty of time has
> gone by that anything like that would gone away long ago.
> 
> We have a lot of aging vbox bugs and I'm trying to get an idea of which ones
> still need work, and which ones have faded away with time.
> 
> Thanks!

Hello! This one is still present with gentoo-sources-3.11.2 and virtualbox-4.2.18.

I updated the postinst script in the meantime, though:
# cat /etc/kernel/postinst.d/00_module-rebuild
---
#!/bin/sh

version="$1" ; shift
image="$1" ; shift

mount-portage.sh

mount -o remount,rw /usr
mount -o remount,rw /opt

#export KERNEL_DIR="/lib/modules/${version}/source"
#export KBUILD_OUTPUT="/lib/modules/${version}/build"
export KERNEL_DIR="/usr/src/linux-${version}"
export KBUILD_OUTPUT="/root/build-${version}"

unset ARCH

emerge @module-rebuild || exit 0
---

If I am not mistaken, this works if I am compiling for the same version as the one that is currently running, but not if ${version} and $(uname -r) disagree.

[Comment 8 Dennis Schridde 2013-10-03 11:10:24 UTC]

Something also seems to messup the kernel source directory, even if KBUILD_OUTPUT!=KERNEL_DIR:
  [...]
  Using /usr/src/linux-3.11.2-gentoo as source for kernel
  /usr/src/linux-3.11.2-gentoo is not clean, please run 'make mrproper'
  in the '/usr/src/linux-3.11.2-gentoo' directory.
  [...]

I then run
# make -C /usr/src/linux-3.11.2-gentoo mrproper
and after running make install in the kernel build directory, they are again "dirty".

[Comment 9 Dennis Schridde 2013-10-03 15:42:26 UTC]

I cross-posted some info at bug #285866 comment #10, especially attachment #360070 [details].

[Comment 10 Ben Kohler 2013-10-03 16:04:16 UTC]

Can you try to reproduce these issues outside of your postinst script?  I can definitely build virtualbox-guest-additions when $(uname -r) doesn't match the KERNEL_DIR version:

# uname -r
3.11.1
ris-ben ~ # KERNEL_DIR=/usr/src/linux-3.11.3 emerge -1 virtualbox-guest-additions
Calculating dependencies... done!
>>> Verifying ebuild manifests
>>> Emerging (1 of 1) app-emulation/virtualbox-guest-additions-4.2.18
>>> Installing (1 of 1) app-emulation/virtualbox-guest-additions-4.2.18
>>> Jobs: 1 of 1 complete                           Load avg: 3.67, 2.76, 1.74
ris-ben ~ #


I think there is a separate problem when KERNEL_DIR != KBUILD_OUTPUT (affecting other vbox pkgs as well), but let's not conflate the 2 issues.

[Comment 11 Dennis Schridde 2013-10-06 14:40:36 UTC]

(In reply to Ben Kohler from comment #10)
> Can you try to reproduce these issues outside of your postinst script?  I
> can definitely build virtualbox-guest-additions when $(uname -r) doesn't
> match the KERNEL_DIR version: [...]
> 
> I think there is a separate problem when KERNEL_DIR != KBUILD_OUTPUT
> (affecting other vbox pkgs as well), but let's not conflate the 2 issues.

Manually executing the postinst script succeeded:
# /etc/kernel/postinst.d/00_module-rebuild 3.11.4-gentoo /boot/vmlinuz-3.11.4-gentoo

Apparently there is something broken during the stage when the postinst scripts are being executed...

[Comment 12 aabarkley3 2014-12-16 16:14:34 UTC]

Time to close this (RESOLVED WON'T FIX). The oldest version in the tree is 4.2.24.