[GRASS-user] [bug #2765] v.buffer bug??

[Hamish]

Bug log:
> http://intevation.de/rt/webrt?serial_num=2765

Hi,

I have a patch I think gets around the famous v.buffer bug (attached),
please test.

I say "gets around" the bug more than solves it, as while I know that
the bug happens when the "sa" segment loops back to "0" and then the
total number of points is set by the final segment number(+2), I don't
really know why, or if this is the problem or the problem is the setting
of npn (total number of points) directly from the seg ID number.
[ lib/vector/Vlib/buffer.c  clean_parallel() ]

So this patch just makes sure that we don't set npn from a smaller "sa"
value than was found in the data.

This fixes the buffer for the test polygon found in the bug log.
I don't know if it fixes the hole-gets-filled-in area problem.


Hamish
-------------- next part --------------
Index: lib/vector/Vlib/buffer.c
===================================================================
RCS file: /home/grass/grassrepository/grass6/lib/vector/Vlib/buffer.c,v
retrieving revision 1.7
diff -u -r1.7 buffer.c
--- lib/vector/Vlib/buffer.c	11 Sep 2006 04:35:07 -0000	1.7
+++ lib/vector/Vlib/buffer.c	1 Oct 2006 08:43:52 -0000
@@ -117,6 +117,7 @@
 void clean_parallel ( struct line_pnts *Points, struct line_pnts *origPoints, double d , int rm_end )
 {
     int i, j, np, npn, sa, sb;
+    int sa_max = 0;
     int first=0, current, last, lcount;
     double *x, *y, px, py, ix, iy;
     static struct line_pnts *sPoints = NULL;
@@ -148,6 +149,10 @@
             G_debug (5, "  current = %d, last = %d, lcount = %d", current, last, lcount);
 	}
 	if ( lcount == 0 ) { break; }   /* loop not found */
+
+	/* ensure sa is monotonically increasing, so npn doesn't reset low */
+	if (sa > sa_max ) sa_max = sa;
+	if (sa < sa_max ) break;
 
 	/* remove loop if in buffer */
 	if ( (sb-sa) == 1 ){ /* neighbouring lines overlap */