Subject: Re: Pentium Bug may cause system crash
To: Dave Huang <>
From: John F. Woods <>
List: current-users
Date: 11/13/1997 00:26:06
> Heh, cool... it works on my NetBSD/mac68k machine too (Centris 660av... an
> 040 machine). Kinda interesting that nobody seems to care about this bug,
> but everyone wants a fix for the Pentium bug (which does affects my
> Pentium).

That's what Intel gets for convincing everyone that Intel's the only game
in town.  Be careful what you wish for, you just might get it.  ;-)

Actually, there was a pretty big stink a while back about the fact that
some revisions of the 68LC040 (or whatever the stripped-down 040 was) didn't
properly take illegal instruction traps on the unimplemented floating point
instructions, thus crippling the Software FPU extension on the Mac; Apple and
Motorola's unhelpful response was that they never promised that Apple never
promised that the 68LC040 *could* do that right, and that if you had used
Apple's SANE library like they tell you to do it wouldn't have been a problem.
(The SANE library figures out at runtime whether to use instructions or
emulation, at only a slight performance penalty; of course, anyone doing
serious floating point isn't INTERESTED in "slight" performance penalties,
so everyone compiles for real instructions instead...)

[Hey, does anyone recall which processor it was that was Formally Verified?
At least, verified as conforming to the specifications, given the usual
waffling about assuming that fussy analog transistors really are perfect
digital switches ;-), with nothing being said about even HOPING to formally
prove the specification correct...  (And, to pretend at least some vague
current-users topicality, has NetBSD been ported to that processor? :-) ) ]

However, I think the Pentium has some unique exposure aspects:  Mac users
(the overwhelming majority of 68K users, I suspect) do not routinely load
tiny 68K applications for execution by their web browsers (sure, they download
large applications all the time, but from my experience, it would be hard
to distinguish an evil Mac application containing an "unlk a7" instruction
from a well-intentioned but badly written one that still manages to lock up
the machine with irritating frequency...).  However, The Software Monopoly
got the brilliant idea of supplanting Java by sending short sequences of
instructions for the processors of The Hardware Monopoly.  (Of course, for
those who used to play the "irritate IE3.0 users" game by including background
sounds, here's the ultimate way to really hose them...)  (No, it is not
comforting that their favored solution is to have users refuse to use ActiveX
applets from sources unable to pay for digital certificates.)