surprisingly, a duplicate case can cause 5c to loop generating
instructions until memory runs out. better to stop cc/pswt.c once
the case is known to be wrong.
it doesn't cause 8c to fail because it doesn't do the direct case
the same way
		while(q->val != v) {
			nextpc();
			p->as = ABCASE;
			patch(p, def);
			v++;
		}
and the subsequent v++ for the next case doesn't expect a matching pair of q->val.

Reference: /n/sources/patch/applied/swt-dup
Date: Mon Nov 30 17:50:48 CET 2009
Signed-off-by: forsyth@terzarima.net

--- /sys/src/cmd/cc/pswt.c	Mon Nov 30 17:46:56 2009
+++ /sys/src/cmd/cc/pswt.c	Mon Nov 30 17:46:54 2009
@@ -18,6 +18,7 @@
 	Case *c;
 	C1 *q, *iq;
 	long def, nc, i, isv;
+	int dup;
 
 	def = 0;
 	nc = 0;
@@ -51,9 +52,14 @@
 	if(debug['W'])
 	for(i=0; i<nc; i++)
 		print("case %2ld: = %.8llux\n", i, (vlong)iq[i].val);
+	dup = 0;
 	for(i=0; i<nc-1; i++)
-		if(iq[i].val == iq[i+1].val)
+		if(iq[i].val == iq[i+1].val) {
 			diag(n, "duplicate cases in switch %lld", (vlong)iq[i].val);
+			dup = 1;
+		}
+	if(dup)
+		return;
 	if(def == 0) {
 		def = breakpc;
 		nbreak++;