ospfv3: fix: re-elect the DR promptly when the DR's adjacency dies of inactivity

When a Full neighbor's inactivity timer fires, OSPFv2 additionally checks
whether the dead neighbor was the Designated Router while this router is the
Backup: if so, it fires NEIGHBOR_CHANGE on the interface so DR re-election
happens immediately rather than on some later Hello. The OSPFv3 port lacked
this step; mirror it.

All 11 ospfv3 + BgpAndOspfv3 fingerprints are byte-identical: in the shipped
examples adjacency teardown is signal-driven (interface down kills neighbors
directly), so the inactivity timer never expires on a Full DR adjacency. The
step matters for silent failures (a DR that stops sending Hellos without a
carrier loss).

Author: avarga

src/inet/routing/ospfv3/neighbor/Ospfv3NeighborStateFull.cc

@@ -34,6 +34,14 @@ void Ospfv3NeighborStateFull::processEvent(Ospfv3Neighbor *neighbor, Ospfv3Neigh
             neighbor->getInterface()->getArea()->getInstance()->getProcess()->setTimer(neighbor->getPollTimer(), neighbor->getInterface()->getPollInterval());
 
         changeState(neighbor, new Ospfv3NeighborStateDown, this);
+
+        // the dead neighbor was the DR and this router is the BDR: re-run the DR election
+        // promptly instead of waiting for the next Hello (mirrors OSPFv2)
+        if (neighbor->getInterface()->getState() == Ospfv3Interface::INTERFACE_STATE_BACKUP &&
+            neighbor->getInterface()->getDesignatedID() == neighbor->getNeighborID())
+        {
+            neighbor->getInterface()->processEvent(Ospfv3Interface::NEIGHBOR_CHANGE_EVENT);
+        }
     }
     else if (event == Ospfv3Neighbor::ONEWAY_RECEIVED) {
         neighbor->reset();