The guest was originally booting Debian's 3.1 kernel. Updating the Guest Additions went fine (after some initial trouble getting the ISO to mount in the first place). Then we wanted to test booting their 3.2.0-4 kernel, and that's where things go wrong.
I knew that the kernel headers would be needed, so I apt-got the "linux-headers-$(uname -r)" package. This evaluated to linux-headers-3.2.0-4-686-pae and -commons.
Mounting the GA ISO and running VBoxLinuxAdditions.run handed back two problems. They don't seem to be directly related to each other.
1) The script reports that "the headers for the current running kernel were not found", because the output of uname -r doesn't match the output of "make kernelrelease". Trying to step through that by hand, it looks like the invocation of "make kernelrelease" might be running into the same problems reported here, as I get the same missing Makefile error when I run the command that VBoxLinuxAdditions.run does. (Running by hand a "fixed" command as suggested in that thread works, but prints a different number, same as discussed at the end of that thread.)
So the warning about the "missing" headers seem to be spurious. The script goes ahead and tries to build the modules, and hits the next problem.
2) Building the video drivers fails. I can't copy the exact text from the guest (no guest additions), nor can I move the log files to a shared folder and open them on the host (no guest additions), so instead you get some screenshots. Giving the user the ability to take screenshots was an excellent idea, whoever thought that up should get free beer.
Here's the early part of vboxadd-install.log:
- Top of the /var/log/vboxadd-install.log file.
- vboxadd-install.png (11.99 KiB) Viewed 1399 times
- DKMS log and source code
- dkms_make_log.png (14.81 KiB) Viewed 1399 times
After getting the failures the first time, I made sure all the vbox* packages were uninstalled and then rebooted to try again. Exact same behavior the second time around.
Booting the previous 3.1 kernel and re-installing the same GA ISO worked fine again, and is running with no problems. We'd like to move to 3.2 eventually, or straight to 3.8 once that makes its way out of Debian experimental, but there's no particular rush here. If there's anything else I can test or info I can provide, I'd be happy to.