Fix: segfault when reading NON_PM empty tin

This is yet another bug due to the failure to realize NON_PM is -1, not
0. Apparently this segfault would only happen when there is a NON_PM tin
which is not spinach, which I initially considered to be an impossible
case, but which turns out is actually possible when tin creation gives
up trying to pick a monster. There is handling for opening and
identifying such "empty" tins, so the "truthful" style of tin label will
now also handle it properly.

Author: copperwater

src/read.c

@@ -292,13 +292,19 @@ tin_text(struct obj *tin, char* buf)
     if (food == HOMEMADE_TIN || !tin_msgs[msgidx]) {
         return (char *) 0;
     }
-    if (tin->tinseed % 5 == 0 && tin->corpsenm) { /* be truthful */
+    if (tin->tinseed % 5 == 0) { /* be truthful */
         msgidx = (tin->tinseed % NUM_MONST_LABELS) + LABEL_MONST_START;
         if (food == SPINACH_TIN) {
             Strcpy(format_arg, "spinach");
-        } else {
+        }
+        else if (tin->corpsenm != NON_PM) {
             Strcpy(format_arg, mons[tin->corpsenm].pmnames[NEUTRAL]);
         }
+        else {
+            /* empty tin; not an impossible case, just very unlikely, when tin
+             * creation fails to pick a tinnable monster 200 times */
+            Strcpy(format_arg, "nothing");
+        }
     }
     else { /* lie */
         if (msgidx == LABEL_BOGUSMON) {