Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!pasteur!ucbvax!ucdavis!uop!exodus
From: exodus@uop.edu (G.Onufer)
Newsgroups: comp.sys.atari.st
Subject: WARNING: Bug (?) in TurboDOS
Message-ID: <1326@uop.edu>
Date: 18 Apr 88 05:59:22 GMT
Organization: UoP-- Sun Lab Administration
Lines: 24
Keywords: turbodos


Well, as much as I love the performance of TurboDOS, I will not use it
until one matter is straightened out.  I just added the original Tandem
drive from my SH204 to the Micropolis drive I had replaced it with.  Now
both are merrily storing all the data I give to them off of the SH204
controller board.  So I thought I would do some timings of TurboDOS/non-Turbo-
DOS.  I copied the TeX sources from SCSI-1/LUN0/PARTf to SCSI-1/LUN1/PARTc
(actually f: to g:) and timed em.  Took 2:34 sec with Beckemeyer's HD accel,
took 2:06 sec with TurboDOS.  Except fsck tells me it's okay the long way and
that there is a free clustor smack in the middle of a file the fast way.
(oh...those should be MINUTES times, not seconds :-)

So, is this a problem with having two drives on one controller?  Depending
on how much TurboDOS assumes about the hardware configuration, it may scrap
the LUN number in its caching.  Time to actually go through the code and
start replacing addresses with labels and adding comments.......Ughhh.

Also, I would not recommend using any disk repair utilities that directly
access the drives while using Turbodos.  May really screw things up if and when
Turbodos flushes its buffers.

Allen:  Any way to force a program such as TurboDOS to flush it's buffers?
Write a desk acc to do it?   (Add symbolic links to the next TOS!!)

G. Onufer