[FIXED] Fix DeliverLastPerSubject consumer policy with interior deletes

server/consumer.go

@@ -4105,6 +4105,7 @@ func (o *consumer) getNextMsg() (*jsPubMsg, uint64, error) {
 			o.updateSkipped(o.sseq)
 		} else {
 			o.lss.seqs = o.lss.seqs[1:]
+			o.sseq = seq
 		}
 		pmsg := getJSPubMsgFromPool()
 		sm, err := o.mset.store.LoadMsg(seq, &pmsg.StoreMsg)
@@ -5353,7 +5354,7 @@ func (o *consumer) selectStartingSeqNo() {
 				if mmp == 1 {
 					o.sseq = state.FirstSeq
 				} else {
-					var filters []string
+					filters := make([]string, 0, len(o.subjf))
 					if o.subjf == nil {
 						filters = append(filters, o.cfg.FilterSubject)
 					} else {

server/jetstream_test.go

@@ -20210,3 +20210,96 @@ func TestJetStreamStreamRetentionUpdatesConsumers(t *testing.T) {
 		})
 	}
 }
+
+func TestJetStreamMaxMsgsPerSubjectAndDeliverLastPerSubject(t *testing.T) {
+	s := RunBasicJetStreamServer(t)
+	defer s.Shutdown()
+
+	nc, js := jsClientConnect(t, s)
+	defer nc.Close()
+
+	const subjects = 300
+	const msgs = subjects * 10
+
+	_, err := js.AddStream(&nats.StreamConfig{
+		Name:              "test",
+		Subjects:          []string{"foo.>", "bar.>"},
+		Retention:         nats.LimitsPolicy,
+		Storage:           nats.FileStorage,
+		MaxMsgsPerSubject: 5,
+	})
+	require_NoError(t, err)
+
+	// First, publish some messages that match the consumer filter. These
+	// are the messages that we expect to consume a subset of. The random
+	// sequences give us interior gaps after MaxMsgsPerSubject has been
+	// enforced which is needed for this test to work.
+	for range msgs {
+		subj := fmt.Sprintf("foo.%d", rand.Intn(subjects))
+		_, err = js.Publish(subj, nil)
+		require_NoError(t, err)
+	}
+
+	// Then publish some messages on a different subject. These won't be
+	// matched by the consumer filter.
+	for range msgs / 10 {
+		subj := fmt.Sprintf("bar.%d", rand.Intn(subjects/10))
+		_, err = js.Publish(subj, nil)
+		require_NoError(t, err)
+	}
+
+	// Add a deliver last per consumer that matches the first batch of
+	// published messages only. We expect at this point that the skiplist
+	// of the consumer will be for foo.> messages only, but the resume
+	// sequence will be the stream last sequence at the time, i.e. the
+	// last bar.> message.
+	_, err = js.AddConsumer("test", &nats.ConsumerConfig{
+		Name:           "test_consumer",
+		FilterSubjects: []string{"foo.>"},
+		DeliverPolicy:  nats.DeliverLastPerSubjectPolicy,
+		AckPolicy:      nats.AckExplicitPolicy,
+	})
+	require_NoError(t, err)
+
+	// Take a copy of the skiplist and the resume value so that we can
+	// make sure we receive all the messages we expect.
+	mset, err := s.globalAccount().lookupStream("test")
+	require_NoError(t, err)
+	o := mset.lookupConsumer("test_consumer")
+	o.mu.RLock()
+	pending := make(map[uint64]struct{}, len(o.lss.seqs))
+	for _, seq := range o.lss.seqs {
+		pending[seq] = struct{}{}
+	}
+	resume := o.lss.resume
+	o.mu.RUnlock()
+
+	// Now fetch the messages from the consumer.
+	ps, err := js.PullSubscribe(_EMPTY_, _EMPTY_, nats.Bind("test", "test_consumer"))
+	require_NoError(t, err)
+	for range subjects {
+		msgs, err := ps.Fetch(1)
+		require_NoError(t, err)
+		for _, msg := range msgs {
+			meta, err := msg.Metadata()
+			require_NoError(t, err)
+			// We must be expecting this sequence and not have seen it already.
+			// Once we've seen it, take it out of the map.
+			_, ok := pending[meta.Sequence.Stream]
+			require_True(t, ok)
+			delete(pending, meta.Sequence.Stream)
+			// Then ack.
+			require_NoError(t, msg.AckSync())
+		}
+	}
+
+	// We should have received every message that was in the skiplist.
+	require_Len(t, len(pending), 0)
+
+	// When we've run out of last sequences per subject, the consumer
+	// should now continue from the resume seq, i.e. the last bar.> message.
+	o.mu.RLock()
+	sseq := o.sseq
+	o.mu.RUnlock()
+	require_Equal(t, sseq, resume+1)
+}