shithub: puzzles

Info  •  Files  •  Log  •  Branches

Kyle Brazell points out that the completion checker considers a
region valid if it has _a_ dot at its centre of symmetry, even if
that dot isn't actually within the region in question.

[originally from svn r7407]

--- a/galaxies.c

+++ b/galaxies.c

@@ -2622,10 +2622,16 @@

      */

     for (i = 0; i < w*h; i++)

         if (sqdata[i].valid) {

-            sqdata[i].cx = sqdata[i].minx + sqdata[i].maxx + 1;

-            sqdata[i].cy = sqdata[i].miny + sqdata[i].maxy + 1;

+            int cx, cy;

+            cx = sqdata[i].cx = sqdata[i].minx + sqdata[i].maxx + 1;

+            cy = sqdata[i].cy = sqdata[i].miny + sqdata[i].maxy + 1;

             if (!(SPACE(state, sqdata[i].cx, sqdata[i].cy).flags & F_DOT))

                 sqdata[i].valid = FALSE;   /* no dot at centre of symmetry */

+            if (dsf_canonify(dsf, (cy-1)/2*w+(cx-1)/2) != i ||

+                dsf_canonify(dsf, (cy)/2*w+(cx-1)/2) != i ||

+                dsf_canonify(dsf, (cy-1)/2*w+(cx)/2) != i ||

+                dsf_canonify(dsf, (cy)/2*w+(cx)/2) != i)

+                sqdata[i].valid = FALSE;   /* dot at cx,cy isn't ours */

             if (SPACE(state, sqdata[i].cx, sqdata[i].cy).flags & F_DOT_BLACK)

                 sqdata[i].colour = 2;

             else