Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!snorkelwacker!apple!well!dsmall
From: dsmall@well.UUCP (David Small)
Newsgroups: comp.sys.atari.st
Subject: Re: Futher Adventures of ATM & Spectre
Message-ID: <15095@well.UUCP>
Date: 19 Dec 89 04:43:58 GMT
References: <22827@brunix.UUCP>
Reply-To: dsmall@well.UUCP (David Small)
Organization: Whole Earth 'Lectronic Link, Sausalito, CA
Lines: 50


My thanks for the very kind comments concerning Adobe Type Manger and Spectre
GCR!

	We have also had many positive responses on other systems. Frankly,
I'm glad I wrote the software the way I did -- no patch files -- so new 
products could run after the software was cast in concrete.

	I did notice two things I might help with. First, we have seen
*some* (not all!!) problems with internal Mega disk drives and GCR. Typically,
the innermost tracks, 64-79, do not read properly. This appears to be due to
horrific electrical noise inside the Mega. Mark Booth (STACE on GENIE) has
extensively commented on how to produce a shield; it takes some thick
aluminum,
like cookie-baking stuff, and a few minutes of time. In 100% of cases, he has
been able
to fix internal drive problems, and he's up to 20 or so drives now.

	Next, MultiFinder is *not* stable, alas, under 2.3K. We have finally
discovered why. As you know, there's a big problem with Mac software storing
into location 0, a program fault. The stores cause a bus error, a crash I
usually recover from. Well, on a hunch, we set the incircuit emulator to
check for *reads* from locations 0 and 4, which do not cause a crash. (See,
on a Mac, 0-7 are read/write RAM except on powerup; on the ST, they are ROM).
Well whoops! Multifinder, through ROM calls, is *reading* location 0, and
getting 602e1e00 (if I remember right), then plugging it into various critical
memory tables. The result is a crash is not far off.

	The solution is to forbid the ROMs from doing that by simple entry
checks for location=0 on these calls. Three were involved if I remember right.

	This greatly, greatly stabilized Multifinder, and to our complete
surprise, MicrosoftWord 3.02. That was the bug in Word (out of memory) that
I had been chasing over a year and failed on. Well, it works now.

	Anyway, I want to caution you on Multifinder. It is not yet stable
on 2.3K. (Note: The "about Spectre" on 2.3K says 2.0 -- our fault.)

	We have fixed this in 2.5a, and ran it extensively at Comdex for
a week with zero crashes, except for one known-multifinder killer D/A.
We would commonly have 5-8 applications going at once.

	If you have a strong need for it, we could probably arrange a
quick-release of 2.5A. It also features on-line configuration (press HELP
and a config page pops up; use F1-F10 to toggle features) and some other
goodies. Doug is also rewriting the launcher for some pretty stunning stuff.

	-- hope this helps,

	-- thanks, Dave / Gadgets