On Tue, Aug 29, 2000 at 10:25:26AM +0930, Greg Lehey wrote:
> At the very least, there appears to be confusion about how to use the
> hints.  I can see two conflicting views here:
> 1.  You must have a /boot/device.hints file, but it may be empty.

This is minimally correct.  I.e. that's what the build system requires.
This works if you build static hints into your kernel.

> 2.  You must have a /boot/device.hints file, and it must contain at
>     least some entries.

This is more correct.  The new world order says that hints are not in the
kernel, instead they are loaded by the loader at boot time.  By default
they are loaded from /boot/device.hints.  Without hints in some form old,
stupid devices don't work.  Unfortunatly, PC consoles are old, stupid
devices for compatability reasons so it's best to have a working hints
file around.

> I ran into this same problem yesterday.  I had noticed it in the cvs
> mailing list, and I found the first entry in UPDATING.  But it didn't
> say what to put in, and I found no other documentation.  Finally John
> Baldwin told me to copy my GENERIC.hints, so I did that, and it
> worked.  But it seems that we should have some documentation here.  On
> the face of it, (1) above seems the most obvious solution.  In that
> case, 'make install' shouldn't fail if there's no device.hints file,
> it should make one.  If it's (2), it can still copy the MYKERNEL.hints
> file.

Installing GENERIC.hints might run the risk of turning the current
anti-footshooting mechinism into a foot shooting device on non-standard
hardware, but I'm kinda inclined to say, who cares since that's a very
far out edge case.  The only problem is that it might annoy those who
insist on static hints.

> Which begs the question: when should the hints file be updated?

When you add new devices which require hints to work.  Since GENERIC.hints
covers the defaults for most of those devices and I don't see too many
new isa drivers going in to the kernel I think that roughly translates to
"never" on modern system and "rarely" on strange systems made of scavenged
parts.  Heck, on most legacy free systems, it really does translate
to never because you can't add anything that would effect drivers that
need hints (device wiring being the only exception I can think of there).

-- Brooks

Any statement of the form "X is the one, true Y" is FALSE.

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message

Reply via email to