*: keep tombstone if revision == compactAtRev

Before this patch, the tombstone can be deleted if its revision is equal
compacted revision. It causes that the watch subscriber won't get this
DELETE event. Based on Compact API[1], we should keep tombstone revision
if it's not less than the compaction revision.

> CompactionRequest compacts the key-value store up to a given revision.
> All superseded keys with a revision less than the compaction revision
> will be removed.

[1]: https://etcd.io/docs/latest/dev-guide/api_reference_v3/

Signed-off-by: Wei Fu <fuweid89@gmail.com>

server/storage/mvcc/hash.go

@@ -63,6 +63,9 @@ func (h *kvHasher) WriteKeyValue(k, v []byte) {
 	if !upper.GreaterThan(kr) {
 		return
 	}
+
+	isTombstone := BytesToBucketKey(k).tombstone
+
 	lower := Revision{Main: h.compactRevision + 1}
 	// skip revisions that are scheduled for deletion
 	// due to compacting; don't skip if there isn't one.
@@ -71,6 +74,17 @@ func (h *kvHasher) WriteKeyValue(k, v []byte) {
 			return
 		}
 	}
+
+	// When performing compaction, if the compacted revision is a
+	// tombstone, older versions (<= 3.5.15 or <= 3.4.33) will delete
+	// the tombstone. But newer versions (> 3.5.15 or > 3.4.33) won't
+	// delete it. So we should skip the tombstone in such cases when
+	// computing the hash to ensure that both older and newer versions
+	// can always generate the same hash values.
+	if kr.Main == h.compactRevision && isTombstone {
+		return
+	}
+
 	h.hash.Write(k)
 	h.hash.Write(v)
 }