Hi,

on Wed, 04 Oct 1995 18:42:15 -0700 Jason Thorpe wrote 
concerning "Re: testing presence of a device (was sun-lamp CVS commits) " 
something like this:

>>> On a NetBSD/sparc machine with a bwtw0 but only one disk (sd0),
>>> [Callisto] 15> sh -c 'if ( < /dev/sd2a ) 2> /dev/null; then echo yes; else echo no; fi'
>>> no
>>> [Callisto] 16> sh -c 'if ( < /dev/bwtwo0 ) 2> /dev/null; then echo yes; else echo no; fi'
>>> yes
>>> [Callisto] 17>

> I apparently missed this original mail...

Hmm ... it was not cc:'ed to you, but you should have received it on the 
list, by now.

> The point is, doing this in shell to present a list like:

> 	You can install on the following devices:
> 		sd0
> 		sd1
> 		sd3

> ...is impractical.  My install.sh for the hp300 basically does:

I don't think so. From a list of possible devices, you can easily determine
which ones are present using a simple loop. (cgd already showed us how ;))
The huge advantage is, that the only arch-dependent part is the list of 
possible devices, which is very easy to "port" ;)

[...]
You wouldn't get duplicate entries, either.

> Also, the technique of opening a device can be problematic in some 
> circumstances ... maybe the disk is there but the partition has the wrong 
> fstype, and thus the driver returns ENXIO?

I don't think that this is a problem, if you specify the raw partition of
a disk, at least it shouldn't, if there is no other problem with that disk.

Maybe we can keep that in mind for the release after 1.1, I don't think we
need to change things now.

Bye, Rolf