incorrect sign when subtracting from sparse(0)

[John W. Eaton]

On  3-Jan-2008, David Bateman wrote:

| David Bateman wrote:
| > Thomas Kasper wrote:
| >> Bug report for Octave 3.0.0 configured for i686-pc-linux-gnu
| >>
| >> Description:
| >> -----------
| >>
| >>   subtraction of a sparse matrix from a sparse scalar with value 0
| >>   results in a wrong sign
| >>
| >> Repeat-By:
| >> ---------
| >>
| >>   octave:1> sparse (0) - sparse (1)
| >>   ans = Compressed Column Sparse (rows = 1, cols = 1, nnz = 1)
| >>
| >>   (1, 1) ->  1
| >>
| >>   % the same for a matrix-valued op2
| >>   octave:2> s = sprand (4,1,.5)
| >>   s =
| >>
| >>   Compressed Column Sparse (rows = 4, cols = 1, nnz = 2)
| >>
| >>     (2, 1) ->  0.96695
| >>     (3, 1) ->  0.26477
| >>
| >>   octave:3> sparse (0) - s
| >>   ans =
| >>
| >>   Compressed Column Sparse (rows = 4, cols = 1, nnz = 2)
| >>
| >>     (2, 1) ->  0.96695
| >>     (3, 1) ->  0.26477
| >>
| > 
| > 
| > I believe the attached patch should fix it, but have to do a complete
| > rebuild of Octave to check it..
| > 
| > D.
| >
| 
| Ok, in fact you also need a patch to MSparse.cc as my previous patch
| only fixes cases like
| 
| s = 1i * sparse(4,1,.5);
| sparse(0) - s
| 
| The attached patch completely fixes the issue for me.

I applied this to cvs trunk and the 3.0 branch.

Thanks.

jwe