On Saturday, 10 Mar 2001, Lourens Veen wrote:
> When running the Sobel edge detect on an image without a selection, the
> topmost scanline of the output is random garbage. From a very quick
> glance at the source it seems that the previous row array isn't
> initialised
Thanks for the good bug report.
It is initialised, but by passing the bogus argument -1 as a y-coord
to the pixel region function, so there's garbage returned.
Here's a patch that fixes this, relative to the sobel.c that's in
1.2.1. Can someone stick this in stable CVS and also merge it with the
current development head. Thanks!
Austin
------------------------------------------------------------
--- sobel.c~ Tue Aug 22 02:26:56 2000
+++ sobel.c Sun Mar 11 14:08:36 2001
@@ -312,10 +312,10 @@
{
gint b;
- if (y == 0)
- gimp_pixel_rgn_get_row (pixel_rgn, data, x, (y + 1), w);
- else if (y == pixel_rgn->h)
- gimp_pixel_rgn_get_row (pixel_rgn, data, x, (y - 1), w);
+ if (y < 0)
+ gimp_pixel_rgn_get_row (pixel_rgn, data, x, 0, w);
+ else if (y >= pixel_rgn->h)
+ gimp_pixel_rgn_get_row (pixel_rgn, data, x, pixel_rgn->h - 1, w);
else
gimp_pixel_rgn_get_row (pixel_rgn, data, x, y, w);
@@ -384,6 +384,7 @@
sobel_prepare_row (&srcPR, pr, x1, y1 - 1, (x2 - x1));
sobel_prepare_row (&srcPR, cr, x1, y1, (x2 - x1));
+
counter =0;
/* loop through the rows, applying the sobel convolution */
for (row = y1; row < y2; row++)
