3.0.2 release (mingw32 check)

[Benjamin Lindner]

John W. Eaton wrote:
> On 25-Aug-2008, Benjamin Lindner wrote:
> 
> | 3)
> |  >>>>> processing 
> | d:\files\admin\octaveforge_svn\trunk\octave-forge\admin\Windows\mingw32\octave\octave-3.0.2\scripts/strings\mat2str.m
> |    ***** assert (mat2str ([-1/3 +i/7; 1/3 -i/7], [4 2]), 
> | "[-0.3333+0i,0+0.14i;0.3333+0i,-0-0.14i]")
> | !!!!! test failed
> | error: assert (mat2str ([-1 / 3, +i / 7; 1 / 3, -i / 7], [4, 
> | 2]),"[-0.3333+0i,0+0.14i;0.3333+0i,-0-0.14i]") expected
> | [-0.3333+0i,0+0.14i;0.3333+0i,-0-0.14i]
> | but got
> | [-0.3333+0i,0+0.14i;0.3333+0i,0-0.14i]
> | 
> | Now this one, I don't consider a bug, rather a stange test.
> | Why would I expect that -i has a real part of -0 (mind the sign!)?
> | Doing
> |    real(-i)
> | yields
> |    ans = 0
> | as expected, and
> |    sprintf("%f", real(-i))
> | yields
> |    0.00000
> | also as expected.
> 
> On my system:
> 
>   octave:1> -i
>   ans = -0 - 1i
>   octave:2> real (-i)
>   ans = -0
> 
> Remember that Octave does not have any pure imaginary data type.  So i
> is stored as (0,1).  Negating that will result in (-0,-1) unless we
> have a special case.  Is that justified?  I'm not sure.  If you want
> to create (0,-1), then you have to write complex(0,-1).
> 
> jwe
> 

On my system

octave:1> -i
ans = 0 - i1
octave:2> real(-i)
ans = 0

I can follow your argumentation, but why does (-i) have a real part of 
(-0) and simultaneously (-1) has a imaginary part of (0) (without the 
sign) ?
The above test expects mat2str([-1/3]) to result in "-0.3333+0i".
Isn't this inconsistent?
1/3 is represented as (1/3,0) and negating it should then yield 
(-1/3,-0). This is what puzzled me about this test.

It looks like an IEEE floating point feature to have -0 ?

Interestingly I cannot create a minus-zero on my system
octave:1> -0
ans = 0
octave:2> -(0)
ans = 0
octave:3> complex(-0,-1)
ans = 0 - 1i

benjamin