From bug-octave-request at bevo dot che dot wisc dot edu  Tue Nov 11 02:11:56 1997
Subject: Unidentified subject!
From: "John W. Eaton" <jwe at bevo dot che dot wisc dot edu>
To: Vincent Cautaerts <vincent at kid2 dot comm dot eng dot osaka-u dot ac dot jp>
cc: bug-octave at bevo dot che dot wisc dot edu
Date: Tue, 11 Nov 1997 02:12:02 -0600

On 11-Nov-1997, Vincent Cautaerts <vincent at kid2 dot comm dot eng dot osaka-u dot ac dot jp> wrote:

| To: bug-octave at bevo dot che dot wisc dot edu
| Subject: poly-2.m test fail : numerical error limit too low ?
| 
| Bug report for Octave 2.1.2 configured for i586-pc-linux-gnu
| 
| Description:
| -----------
| 
| The test "poly-2.m" fails
| 
| poly([1,2;3,4])-[1,-5,-2]
| 
| returns [0 0 2.4e-15]
| 
| The complete test is
| all (all (poly ([1, 2; 3, 4]) - [1, -5, -2] < 2 * eps))
| 
| and fail because eps=2.2*10-16
| 
| Repeat-By:
| ---------
| 
| make check
| 
| Fix:
| ---
| 
| Change the test to:
| 
| all (all (poly ([1, 2; 3, 4]) - [1, -5, -2] < 20 * eps))

I've already changed this test to be:

  all (all (abs (poly ([1, 2; 3, 4]) - [1, -5, -2]) < sqrt (eps)))

I think that should be an acceptable limit.

Thanks,

jwe