I got my 1.3 Alpha CD on Saturday and have been installing it over the
weekend. Everything seems to be okay apart from a few things.

It still won't recognise my CD-Rom drive (Toshiba XM-?5307?, quad speed
ATAPI). On bootup it reports finding atapibus0 and wd0 (my original
Conner 850MB disk) but not wd1 or cd0 (or what ever the CD should come
out as).

Despite this I managed to install 1.3a without problem. First onto a
borrowed SCSI disk connected to a Powertec SCSI II (a 4.3GB UltraWide
Seagate Barracuda, gets lots a MB/s on a PC, over 5MB/s under RiscOS,
max 150k/s under BSD!). Then onto the Conner having copied the original
RiscBSD partition from the Conner to the Barracuda.

This was done using UnixFS from RiscOS to transfer the distrib directory
onto the BSD partition. I originally had the Conner as 400MB RiscOS &
400MB BSD but now I have the whole disk as BSD. Likewise, the Barracuda
is pure BSD so no partition table worries [Mr. Nelson, even if 1.3a does
not work with Powertec partitions, it is quite happy to use whole disk
without a table but be warning, transfer rates are still pitiful].

I am now in the process of transferring my original installation back
across to the new one (work, home space, configuration, etc).
Unfortunately, the kernel keeps panicking. It always seems to occur when
doing large disk i/o tasks. But that may simply be that all I have been
doing so far is tar/untar-ing from one disk to the other.

The actual panic seems to change (page permission fault, translation
fault, summat else but I forget what) as does the user land program
which caused it (find, tar, etc.). But the instruction is always ldmdb
r?, r?, {r? - r?}. I will write it down next time it happens but the
last crash was late last night and so I just went to bed instead! I can
also supply a kernel core dump (50MB) if it would be helpful.

BTW, in response to the call of people who want fast SCSI drivers should
stop whining and start coding. I am too busy coding my own projects to
attempt kernel land drivers. I am happy to test any code thrown at me
though (I have a removable SCSI drive [MagOpt] so I have a SCSI device I
could live with being trashed).

One last point, I attempted a rebuild of the full source tree (cd
/usr/src ; make) and it make it quite a long way but then stopped with a
compiler error. I'm not sure where though. I only set it going as
something to do whilst watching a film :-). It managed over an hour of
compilation without panicking though.

Oh yes, yet another last point. The documentation says that bb_trash
should be run to remove any RiscOS partition info from a disk before
attempting a complete BSD only install. This program does not seem to
exist. 'cat /dev/zero > /dev/wd0c' works fine though :-).


John Harrison
R&D, 3Dlabs