Gentlepeople,
In gimp 1.1.9 (Checkout Friday Sep 3, 1999) the conversion of a
selection to a selection border ("right mouse button
menu"->Select->Border) renders a five pixel border regardless of the
content of "Border selection by:"
The handler of the OK button event appears to be
static void gimage_mask_border_callback (GtkWidget *widget, gpointer
client_data, gpointer call_data)
implemented in gimp/app/commands.c (CVS-1.74 Fri Sep 3 01:21:12 1999
Line 1327). It casts call_data, reflecting the content of the dialog
box, into a static gint selection_border_radius. A few lines further,
it initializes local automatic variables radius_x and radius_y:
1343: radius_x = radius_y = selection_feather_radius;
and passes these down the food chain to gimage_mask_border().
Methinks this initializer has been inadvertantly used instead of
selection_border_radius, which appears to be now decoupled from
computations. Shouldn't this be something like:
1343:' radius_x = radius_y = ((gdouble)selection_border_radius);
Or is some weighting of selection_feather_radius also a term in this
sum to fatten the border a smidgen prior to a feathering convolution?
Patch to commands.c implementing a 'non-fattening version' of this
amendment is attached.
Garry Osgood
--- commands.c.old Sun Sep 5 22:00:19 1999
+++ commands.c Sun Sep 5 21:49:42 1999
@@ -1340,7 +1340,9 @@
g_free (call_data);
unit = (GUnit) gtk_object_get_data (GTK_OBJECT (widget),
"size_query_unit");
- radius_x = radius_y = selection_feather_radius;
+ /* radius_x = radius_y = selection_feather_radius; gosgood@xxxxxxx */
+
+ radius_x = radius_y = (gdouble)selection_border_radius;
if (unit != UNIT_PIXEL)
{
