Hi,

on Wed, 04 Oct 1995 01:34:15 -0700 Jason Thorpe wrote 
concerning "Re: sun-lamp CVS commits " something like this:

> It could be easily adapted to other ports, yes.  I made an effort to keep 
> machine-dependent goo to a minimum.  Probably the most machine-dependent 
> part in the installation script is the regexp's applied to dmesg output 
> to determine which devices are there (not much else you can do in a shell 
> script :-).

This is not only machine dependent, it may even be problematic on machines
with a long uptime, because the boot messages may not be there anymore.
(On the other hand, on the Sparc the message buffer is not destroyed across
multiple boots, so there may be messages from more than one boot in the
buffer.)

I think, if you want to test the presence of a device, you should try to
open it for reading. If you succeed, the device is present. Unfortunately
you're right, this is not really possible to do with sh. Maybe we could
enhance test to do that?

Just some thoughts.

Bye, Rolf