Add Storage.Term() method and hide the first entry from other methods.

The first entry in the log is a dummy which is used for matchTerm
but may not have an actual payload. This change permits Storage
implementations to treat this term value specially instead of
storing it as a dummy Entry.

Storage.FirstIndex() no longer includes the term-only entry.

This reverses a recent decision to create entry zero as initially
unstable; Storage implementations are now required to make
Term(0) == 0 and the first unstable entry is now index 1.
stableTo(0) is no longer allowed.

etcdserver/server_test.go

@@ -840,9 +840,29 @@ func TestSnapshot(t *testing.T) {
 	n := raft.StartNode(0xBAD0, mustMakePeerSlice(t, 0xBAD0), 10, 1, s)
 	defer n.Stop()
 
-	// Save the initial state to storage so we have something to snapshot.
+	// Progress the node to the point where it has something to snapshot.
+	// TODO(bdarnell): this could be improved with changes in the raft internals.
+	// First, we must apply the initial conf changes so we can have an election.
 	rd := <-n.Ready()
 	s.Append(rd.Entries)
+	for _, e := range rd.CommittedEntries {
+		if e.Type == raftpb.EntryConfChange {
+			var cc raftpb.ConfChange
+			err := cc.Unmarshal(e.Data)
+			if err != nil {
+				t.Fatal(err)
+			}
+			n.ApplyConfChange(cc)
+		}
+	}
+	n.Advance()
+
+	// Now we can have an election and persist the rest of the log.
+	// This causes HardState.Commit to advance. HardState.Commit must
+	// be > 0 to snapshot.
+	n.Campaign(context.Background())
+	rd = <-n.Ready()
+	s.Append(rd.Entries)
 	n.Advance()
 
 	st := &storeRecorder{}
@@ -854,7 +874,7 @@ func TestSnapshot(t *testing.T) {
 		raftStorage: s,
 	}
 
-	srv.snapshot(0, []uint64{1})
+	srv.snapshot(1, []uint64{1})
 	gaction := st.Action()
 	if len(gaction) != 1 {
 		t.Fatalf("len(action) = %d, want 1", len(gaction))