David Laight <dsl@netbsd.org> writes:
> Log Message:
> If we are builging a small kernel [1], don't inline all these functions.
> Saves over 2k and lets i386 rescue_tiny build again.
> [1] if MALLOC_NOINLINE is defined - not ideal but...

I honestly don't think this constant battle to make 'rescue_tiny' keep
working is worth it at all. Really.

At some point, we bite the bullet, and move away from
ramdisk-in-kernel boot media on the i386. I introduced them, and I
think it is time they died.

Why did we introduce the ramdisk-in-kernel boot stuff? We did it to
eliminate a reboot during installs. We lived in a world ten years ago
where you couldn't be sure people had a CD or a network, and there
were no disk-on-key units. We had install packages available in the
form of sliced up floppy sized units because people might have no
other way to do stuff!

However, that world is gone. There is no point any more. You can now
just put a root image on your CD or disk-on-key or over NFS and be
done. This is even more true for AMD64 -- there is no AMD64 box that's
10 or 15 years old to need legacy support.

We should quit with this stupid bit fiddling and change the whole
mechanism and be done with it. We aren't helping anyone by searching
for ways to keep 'rescue_tiny' building in the first place.

Perry