[ᚬmaster] 🎉 Rc/v0.14.0

[doitian]

This is the release for Rylai V3. The testnet will reset and deploy this version on Jun 15.

---

Features

• feat: Peers handle disconnect #908: Peers handle disconnect (@keroro520)

  • Handle disconnected inside Peers
  • Rename n_sync_started/n_protected_outbound_peers
  • Remove imlicitly clone for Peers

• feat: secp256k1 multisig #891: Secp256k1 multisig (@jjyr)

  Used in alert system feat: Alert system #887

• feat: limit TXO set memory usage #845: Limit TXO set memory usage (@yangby-cryptape)

  This is a breaking change: b:database

  Commits

  • feat: add a function to traverse all keys for database

  • feat: save cell set into database (but not used)

    BREAKING CHANGE: Add new column to store cell set into database.

  • refactor: loading CellSet do not have to load all blocks

  • perf: delete transactions whose all cells are dead

  • fix: the order to apply cell set diff is very important

  Issue

  Limit TXO set memory usage #228

  Notes

  • I insist on the opinion: save live cells is better than save transactions which include at least one live cell.

    We can use a only-key-and-no-value database (for example, Redis) to store live cells (CellOutPoint as key).

    I tried, but it's too difficult to adapt to current logic, so I gave up.

  • I didn't change current design about how to use CellSet, I just made them be persisted.

  • All transactions which include at least one live cell, are still kept in memory.

    But the transactions which don't have any live cells are removed from memory.

• feat: revise uncle rule #874: Revise uncle rule (@zhangsoledad)

  This is a breaking change: b:consensus, b:database

  1. get rid uncle age limit
  2. try include disconnected block as uncle

• feat: tweak consensus params #920: Tweak consensus params (@zhangsoledad)

  This is a breaking change: b:consensus

  tweak block_cycles_limit and min_block_interval

• feat: wrap the log macros to fix ill formed macros  #897: Wrap the log macros to fix ill formed macros (@yangby-cryptape)

  Commits:

  • chore: update the log filter template in configuration
  • chore: apply new log macros to sync & relay
  • chore: apply new log macros to util/* & ckb-bin
  • chore: apply new log macros to verification & shared
  • chore: apply new log macros to rpc & notify
  • chore: apply new log macros to network
  • chore: apply new log macros to miner
  • chore: apply new log macros to chain
  • (important) feat: wrap the log macros to fix ill formed macros

  And, we have to update the log filters, add prefix ckb- for all our crates, since I use target: env!("CARGO_PKG_NAME").

• feat: synchronizer and relayer share BlocksInflight #919: Synchronizer and relayer share BlocksInflight (@keroro520)

  • test: Add test for losing BlockTXN messages

  • fix: Remove relayer's compact_block_filter, filter duplicated compact blocks via BlocksInFlight
    Currently, we discard the new arriving compact blocks if we had already received the same one, via RelayState::compact_block_filter. The purpose is avoiding re-processing/re-requesting the same compact blocks. But the problem is, if we failed to reconstruct a compact block, we cannot process this compact block anymore, before receiving the corresponding transactions from peer1. That's terrible if peer1's inventory was lost on the way.

    Here I remove RelayState::compact_block_filter, and insteadly use Peers::blocks_inflight to avoid re-requesting to same peers.

  • fix: Cache pending compact block at peer-level, but not global-level

• feat: add a transaction error InsufficientCellCapacity #924: Add a transaction error InsufficientCellCapacity (@yangby-cryptape)

  For Do not reuse the error CapacityOverflow #900.

• feat: make a better error message for miner when method not found #926: Make a better error message for miner when method not found (@yangby-cryptape)

  For Optimize error message on rpc call get_block_template MethodNotFound #904.

Bug Fixes

• fix: calculate the current median time from tip #878: Calculate the current median time from tip (@keroro520)

  This is a breaking change: b:consensus

  BREAKING CHANGE: Original implementation use [Tip-BlockMedianCount .. Tip-1] to calculate the current block median time. According to the notion of BlockMedianTime in bip-0113 , here change to use [Tip-BlockMedianCount+1 .. Tip] instead.

• fix: Sync blocked by protected peer #915: Sync blocked by protected peer (@TheWaWaR)

• fix: proposal table reload #906: Proposal table reload (@zhangsoledad)

• fix: Cuckoo cycle verification bug (@yangby-cryptape)

Improvements

• refactor: move best_known_headers/last_common_headers into peer_state #909: Move best_known_headers/last_common_headers into peer_state (@keroro520)

  • best_known_header and last_common_header are both state of peer. So I move them into PeerState. And add wrap functions for them, so that hides the locks outside.

Misc

• New verification model #913: New verification model (@xxuejie)

  This is a breaking change: b:consensus, b:database, b:p2p, b:rpc

  Based on feedbacks gathered earlier, we are revising our verification
  model with the following changes:

  • When validating a transaction, CKB will grab all lock scripts from
    all inputs, and group them based on lock script hash. The script in
    each group will only be run once. The lock script itself will have
    to do the validation task for all inputs in the same group
  • CKB will also grab all type scripts from inputs and outputs(notice
    different from previous version, the type scripts in inputs are
    included here as well), and group them based on type script hash as
    well. Each type script in each group will also be run once. The type
    script itself needs to handle the validation task within the group
  • Syscalls are also revised to allow fetching all the
    inputs/outputs/witnesses within a single group.
  • Input args is removed since the functionality can be replicated elsewhere

[nervos-bot]

@xxuejie is assigned as the chief reviewer

[doitian]

Close because of the wrong source repo