Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!cs.utexas.edu!ut-emx!bill
From: bill@ut-emx.UUCP (Bill Jefferys)
Newsgroups: comp.sys.mac
Subject: Re: Bug in EXCEL v. 2.2, and a flame
Keywords: Excel
Message-ID: <16495@ut-emx.UUCP>
Date: 2 Aug 89 19:20:27 GMT
References: <4134@utastro.UUCP> <5680@tekgvs.LABS.TEK.COM>
Reply-To: bill@emx.UUCP (Bill Jefferys)
Organization: UTexas Computation Center, Austin, Texas
Lines: 36

In article <5680@tekgvs.LABS.TEK.COM> larryh@tekgvs.LABS.TEK.COM (Larry Hutchinson) writes:
#
#I am not convinced that this is a bug at all.
#
#Directly using either the 68882 OR using SANE (which probably calls the 882)
#I get:
#
#sqrt(0.999975)=
#    0.999987499921874
#
#as a check on the answer:
#0.999987499921874 * 0.999987499921874=
#    0.9999749999999998
#which is not too shabby.
#
#Using the value you state, I get
#0.9999749992187 * 0.9999749992187=
#    0.9999499990624391
#which is obviously far less accurate.
#
#
#This is on an SE/30.  But if this (and your Mac+) is the correct answer
#then that implys your Mac II gave the wrong answer!  Pretty wierd stuff.
#(My HP15C gives the same results to 10 digits.)
#
#Where did you get your value?  How did you check it?

Typing error. The Mac II gives 0.99998749992187. I dropped the 8
when copying the number.

An SE/30 of course has the coprocessor & can be expected to give
the correct answer. On the plus, you get 0.9999875 exactly. This
is the wrong answer, and the bug. YOU WILL NOT BE ABLE TO REPRODUCE
THIS BUG ON A MACHINE WITH A COPROCESSOR.

Bill Jefferys