[AD] Fix overflow in create_light_table()

[Eric Botcazou <ebotcazou@xxxxxxxxxx>]

I came accross this thread on allegro.cc

	http://www.allegro.cc/forums/view_thread.php?_id=275499&request=1055944986&;

where Johan Peitz reports a problem with create_light_table().

It turns out that the interpolation formula used in the function can overflow 
for x=0, r=g=b=63 and pal[y].r=pal[y].g=pal[y].b=63. We end up with an index 
that is outside the required 5 bits for rgb_map->data.

Fortunately, there is the fix by Martijn van Lersel in the same thread. It 
simply restores the symmetry between the cases x=0 and x=255, so that we are 
now guaranteed to stay within the 5 bits.

I reworked it a little and applied it to trunk and branch.

-- 
Eric Botcazou

--- /home/eric/cvs/allegro/src/color.c	Sat Jun  7 10:34:40 2003
+++ allegro/src/color.c	Thu Jun 19 08:38:54 2003
@@ -665,20 +665,21 @@
 void create_light_table(COLOR_MAP *table, AL_CONST PALETTE pal, int r, int g, int b, void (*callback)(int pos))
 {
    int r1, g1, b1, r2, g2, b2, x, y;
-   unsigned int tmp;
+   unsigned int t1, t2;
 
    if (rgb_map) {
       for (x=0; x<PAL_SIZE-1; x++) {
-	 tmp = (255 - x) * 65793;
-	 r1 = r * tmp + (1 << 24);
-	 g1 = g * tmp + (1 << 24);
-	 b1 = b * tmp + (1 << 24);
+	 t1 = x * 0x010101;
+	 t2 = 0xFFFFFF - t1;
+
+	 r1 = (1 << 24) + r * t2;
+	 g1 = (1 << 24) + g * t2;
+	 b1 = (1 << 24) + b * t2;
 
-	 tmp = (1 << 24) - tmp;
 	 for (y=0; y<PAL_SIZE; y++) {
-	    r2 = (r1 + (unsigned int)pal[y].r * tmp) >> 25;
-	    g2 = (g1 + (unsigned int)pal[y].g * tmp) >> 25;
-	    b2 = (b1 + (unsigned int)pal[y].b * tmp) >> 25;
+	    r2 = (r1 + pal[y].r * t1) >> 25;
+	    g2 = (g1 + pal[y].g * t1) >> 25;
+	    b2 = (b1 + pal[y].b * t1) >> 25;
 
 	    table->data[x][y] = rgb_map->data[r2][g2][b2];
 	 }
@@ -688,16 +689,17 @@
    }
    else {
       for (x=0; x<PAL_SIZE-1; x++) {
-	 tmp = (255 - x) * 65793;
-	 r1 = r * tmp + (1 << 23);
-	 g1 = g * tmp + (1 << 23);
-	 b1 = b * tmp + (1 << 23);
+	 t1 = x * 0x010101;
+	 t2 = 0xFFFFFF - t1;
+
+	 r1 = (1 << 23) + r * t2;
+	 g1 = (1 << 23) + g * t2;
+	 b1 = (1 << 23) + b * t2;
 
-	 tmp = (1 << 24) - tmp;
 	 for (y=0; y<PAL_SIZE; y++) {
-	    r2 = (r1 + (unsigned int)pal[y].r * tmp) >> 24;
-	    g2 = (g1 + (unsigned int)pal[y].g * tmp) >> 24;
-	    b2 = (b1 + (unsigned int)pal[y].b * tmp) >> 24;
+	    r2 = (r1 + pal[y].r * t1) >> 24;
+	    g2 = (g1 + pal[y].g * t1) >> 24;
+	    b2 = (b1 + pal[y].b * t1) >> 24;
 
 	    table->data[x][y] = bestfit_color(pal, r2, g2, b2);
 	 }