> Yes it was good that it failed since my bios is setup to pxeboot if the
> hard drive doesn't boot.

Only works because your bios treats 'int 0x18' as a 'boot failed'
entry.  Many do, but it is actually the entry to rom basic for a
cassette tape boot (or something like that).

> > I'll think about that one, instalboot doesn't do that for any other
> > architecture.
> 
> Well, I can understand not putting hacks in for one arch, but since 
> installboot already has to know about i386 specifically, I don't see why
> an extra check for the sanity of /boot would hurt.

The actual problem in that the code doesn't know where /boot is!
You could quite easily be installing a bootstrap into other than the
current root partition (you can boot alternate root fs that way).

The code would have to use the stuff that grovels through the filesystem
structures (normally used to get a list of fs blocks) to read the /boot
file.  The changes require structural changes to installboot.

> At the very minimum the man page needs updating since it says to copy
> boot.MACHINE to /boot.

Is that correct for any system?
Do any actually look for /boot.MACHINE then /boot?
Is there space in any of the bootstraps to code the latter?

> > I've not looked at pxeboot, really needs building from the new sources.
> > (but I can't test it)
> 
> It builds and works from the latest -current as of this morning.
> The issue with 'quit' has been there forever, netboot had it also.
> 
> I'll  try to remember to look into making a reset on quit patch.

Add 'int $0x18' before the halt/loopstop.


	David

-- 
David Laight: david@l8s.co.uk