This page tracks releases and changes for v0.39.0. For the full release history, see the CHANGELOG on GitHub.
BUG FIXES
[evidence]Add validation for Light Client Attack evidence ByzantineValidators[types]Fix buffer offset bug inProposerPriorityHashthat caused hash collisions when validator priorities differed[p2p]fix(privval): Ephemeral Port Exhaustion[blocksync]fix(blocksync):ExtendedCommitverification via next blocksLastCommit- [p2p] fix(lp2p): enforce stream max size (#5647)
[metrics]fix(metrics)!: peer_send_queue_size[statesync]fix adaptive_sync and streamline stateSync logic[blocksync]Modify blocksync to use full commit verification instead of light[adaptivesync]Simplify loop, reuse blockExec.ValidateBlock
IMPROVEMENTS
[ci]: add lp2p testnet (#5643)[mempool]feat!(p2p): introduce follower-mode. Improve lib-p2p integraap access[types]Add validation forAuthorityParams.Authorityfield in consensus params, enforcing a maximum length of 256 characters (#5511)[mempool]perf(mempool/cache): Optimize LRUTxCache.Remove to reduce lock contention and map access[e2e]add support for testing different keytypes, including BLS[crypto]Reduce BLS signature size to 48 bytes by increasing pubkey size to[statesync]Add configurablemax-snapshot-chunksparameter to validate max amount of chunks in aSnapshotResponse.[p2p]feat(lp2p): make reactor queue configurable[cli]print lib-p2p peer id[p2p]Add warning when go-libp2p transport is enabled, conveying that the setting[p2p]feat(p2p): add adaptive sync for comet-p2p
FEATURES
[p2p]feat(lp2p): implemented resource limiter (#5671)[p2p]feat(consensus): add adaptive sync blocksync-to-consensus ingestion (#5633)[p2p]feat(lp2p): implement Peer info methods (NodeInfo,RemoteIP,RemoteAddr,IsOutbound)[p2p]feat(lp2p): stop/reconnect peers that failed (#5618)[p2p]Add experimental support for lib-p2p networking (#5463)[crypto]Add support for BLS12-381 keys. Since the implementation needs[mempool]Add a metric (a counter) to measure whether a tx was received more than once.[p2p]RenameIPeerSet#ListtoCopy, addRandom,ForEachmethods.[mempool]When the node is performing block sync or state sync, the mempool- Optimized the PSQL indexer
[p2p]makePeerSet.Removemore efficient (Author: @odeke-em)[light]Remove duplicated signature checks inlight.VerifyNonAdjacent[state/indexer]Lower the heap allocation of transaction searches[libs/json]Lower the memory overhead of JSON encoding by using JSON encoders internally[log]allow strip out all debug-level code from the binary at compile time using build flags[types]Small reduction in memory allocation via swapping Key with Equals in VoteSet[event-bus]Remove the debug logs in PublishEventTx, which were noticed production slowdowns.[state/execution]Cache the block hash computation inside of the Block Type, so we only compute it once.[consensus/state]Remove a redundantVerifyBlockcall inFinalizeCommit[p2p/channel]SpeedupProtoIOwriter creation time, and thereby speedup channel writing by 5%.[p2p/conn]Minor speedup (3%) to connection.WritePacketMsgTo, by removing MinInt calls.[blockstore]Remove a redundantHeader.ValidateBasiccall inLoadBlockMeta, 75% reducing this time.[p2p]Lowerflush_throttle_timeoutto 10ms[types]Significantly speedup types.MakePartSet and types.AddPart, which are used in creating a block proposal[types] Make a new methodGetByAddressMutforValSet`, which does not copy the returned validator.[consensus]Make Vote messages only take one peerstate mutex[consensus]Make the consensus reactor no longer have packets on receive take the consensus lock. Consensus will now update the reactor’s view after every relevant change through the existing synchronous event bus subscription.[p2p/conn]Speedup secret connection large writes, by buffering the write to the underlying connection.[consensus]Make broadcastingHasVoteandHasProposalBlockPartcontrol messages useTrySendinstead ofSend. This saves notable amounts of performance, while at the same time those messages are for preventing redundancy, not critical, and may be dropped without risks for the protocol.[p2p/conn]Removes several heap allocations per packet send, stemming from how we double-wrap packets prior to proto marshalling them in the connection layer. This change reduces the memory overhead and speeds up the code.[p2p/conn]Speedup secret connection large packet reads, by buffering the read to the underlying connection.[mempool]In the broadcast routine, get the pointer to the peer’s state once, before starting to iterate through the list of transactions.[consensus]Make mempool updates asynchronous from consensus Commit’s,- [consensus] Add peer height metric publication to the consensus reactor’s peer state.
BUG-FIXES
[evidence]Use structured logging for consensus buffer flush error[consensus]Reject oversized proposals[store]Prune extended commits properly[mempool]Fix mutex inCListMempool.Flushmethod, by changing it from read-lock to write-lock[crypto/bls12381]Fix JSON marshal of private key[crypto/bls12381]ModifySign,Verifyto usedstMinPk[bits]Validate BitArray mismatched Bits and Elems length[cli]Prevent inadvertent rollover of IPs incometbft testnetconfig generator
API-BREAKING
[p2p]RenameIPeerSet#ListtoCopy, addRandom,ForEachmethods.[crypto]Remove Sr25519 curve[rpc]The endpointsbroadcast_tx_*now return an error when the node is