I have several systems here. I build three different kernels: one supporting more than 4GB RAM, one supporting up to 4 GB, and the other is heavily stripped kernel for some old PI and PII machines.

In the past I used the CONFIG_LOCALVERSION option to differentiate the kernels. Yet now I wonder whether this is a practical idea.

A problem arises when I want to compile updated packages that build kernel modules, such as nvidia, lirc, virtualbox, etc. The nvidia build seems particularly anal about embedded kernel version numbers.

As the kernel version is the same for each custom build other than the config file options, I would like to build generic packages that install on any machine and avoid building packages unique for each kernel build. Seems I need to avoid embedded version numbers to do that.

Is the solution simply to stop using CONFIG_LOCALVERSION and manually rename the individual kernels after compiling?

Probably. I've never rarely used it.

Hi Woodsman - How about using the EXTRAVERSION string in the top level Makefile?

VERSION = 2
PATCHLEVEL = 6
SUBLEVEL = 37
EXTRAVERSION = your custom name that identifies this machine
NAME = Flesh-Eating Bats with Fangs

Noted. I'll try that next time I compile a kernel and some packages that create kernel modules. Hopefully I remember to post the results here.