On Mon, 11 Dec 2000, Todd Vierling wrote:

# : a) Add `-DKERNEL_foo' to the IDENT line, where `foo' is the kernel
# :    config file name. I would find this useful when I'm adding hacks
# :    or debugs for a specific machine which has its own config file
# :    (e.g, my laptop with semi-borken APM), so I can wrap system
# :    specific code with #ifdef KERNEL_foo
# 
# Be aware that config(8) can config kernels by a method other than "filename
# == kernelname" -- you can also create a directory with a file called CONFIG
# in it (usually with a "source /sys" directive in it) and just run config(8).
# I do this all the time to build kernels completely outside of the syssrc
# tree.  In this case, the kernelname should probably be taken from the last
# directory component.

Am I reading this right?  I think you might be answering the wrong question,
here.  "kernelname" typically == "netbsd", while "filename" usually ==
"MACHINE_NAME_IN_CAPS" or some other moniker.  So, yes, you're right, the
two differ. :-).

If I misinterpreted, my apologies...

...ALso, I just discovered that the 'ident' clause actually works, which
is nice.  I can build two slightly different kernels from the same tree
and have the (MACHINE) change (which saves time, and I can tell which one
is which).  Thanks to whomever implemented that code, because it DIDN'T
do that for the longest time, as far as I could tell (unless, again,
I missed something).

# Don't create this `feature' on a.out; it'll give more incentive for
# platforms to switch to ELF.  <g>

				--*greywolf;
--
*BSD: Twice the Bits-Clean of other Leading OSes.