From webb@groucho.csl.uiuc.edu Fri Mar  8 16:38:35 1991
Received: from watmath.waterloo.edu by orion.waterloo.edu with SMTP
	id <AA08039>; Fri, 8 Mar 91 16:38:31 EST
Received: from surfer.EPM.ORNL.GOV by watmath.waterloo.edu with SMTP
	id <AA14536>; Fri, 8 Mar 91 16:38:23 EST
Received: by surfer.EPM.ORNL.GOV (5.61/1.34)
	id AA10786; Fri, 8 Mar 91 16:38:12 -0500
Received: from patience.Stanford.EDU by surfer.EPM.ORNL.GOV (5.61/1.34)
	id AA10776; Fri, 8 Mar 91 16:37:43 -0500
Received: from groucho.csl.uiuc.edu by patience.Stanford.EDU (4.0/inc-1.5)
	id AA03771; Fri, 8 Mar 91 13:37:49 PST
Message-Id: <9103082137.AA03771@patience.Stanford.EDU>
Received: by groucho.csl.uiuc.edu
	(15.11/15.6) id AA08087; Fri, 8 Mar 91 15:35:38 cst
From: Jennifer Webb <webb@groucho.csl.uiuc.edu>
Subject: bugs in simplex matlab routine from netlib
To: na.wolkowicz@patience.Stanford.EDU
Date: Fri, 8 Mar 91 15:35:35 CST
Mailer: Elm [revision: 64.9]
Status: RO

I have been using the simplex1 m-files which you wrote, having obtained them
through netlib.  They have been a tremendous help!  However, I believe that
my test cases have uncovered two errors in the loop following the comment
"Ensure that no artificial variable remains in the basis" in main.m. 

First of all, in order to avoid inconsistent array dimensions later on,
(c=c-ap(i,:)*ratio computation), I added a line immediately before the
"Ensure..." comment:
                        c=c(1:n1+m1);

The other change was in adding an absolute value to the "%pivot column"
line, ie.
                        t=n1m1to1(abs(a(i,:))>eps1);

Thank you again for the use of your routine.

                                Sincerely,

                                Jennifer Webb