[OctDev] Re: Off-by-one error in bmpwrite.m?

[Paul K. <pki...@us...>]

On Jul 5, 2004, at 6:28 AM, Per Persson wrote:

> Hi,
> in preparing a new image.m for Mac OS X (based on code from J. Koski)  
> that uses bmpwrite(), and I think I've found a bug in bmpwrite.m.
>
> The color for {R,G,B}={1, 1, 1} is random on OS X and I think the  
> problem is that the index into colormap in a .bmp file is zero-based  
> (couldn't find a definitive authority). The help for bmpwrite doesn't  
> state that a zero-based index is used, so I assume that the patch  
> below is the right fix.

Here's a pretty good authority that says you are right:

	http://www.dcs.ed.ac.uk/home/mxr/gfx/2d-hi.html

Please update the server.

Thanks,

Paul Kienzle
pki...@us...

>
> /Per
>
> Index: main/image/bmpwrite.m
> ===================================================================
> RCS file: /cvsroot/octave/octave-forge/main/image/bmpwrite.m,v
> retrieving revision 1.2
> diff -u -d -b -w -r1.2 bmpwrite.m
> --- main/image/bmpwrite.m       27 Nov 2002 08:40:10 -0000      1.2
> +++ main/image/bmpwrite.m       5 Jul 2004 10:16:01 -0000
> @@ -37,7 +37,7 @@
>
>      ## raster image, each line on a 32-bit boundary, padded with zeros
>      ## lines written bottom to top.
> -     
> fwrite(file,postpad(flipud(x)',ceil(columns(x)/4)*4),"uchar",0,arch);
> +     
> fwrite(file,postpad(flipud(x-1)',ceil(columns(x)/ 
> 4)*4),"uchar",0,arch);
>      fclose(file);
>  endfunction
>