Skip to content

Instantly share code, notes, and snippets.

What would you like to do?
- Core
- Switched our testnet1 to mainnet.
- Perf improvement to do, in particular repeated hashTreeRoot
- Stack overflow as the beacon state is huge
- Update/rewind states to apply state transition on blocks coming from the network
-> concern about DOS vector
- Networking
- more reliable syncing
- Debugging Discovery and Libp2p
- Memory leaks in networking/streams: solved one in libp2p, another to track down
- Auditing:
- removing stale code, improving handling of errors
- Interop:
- Fixed a bug of different genesis between Nimbus and Lighthouse
- Discovery is working but can’t talk yet
- Tried to run Prysm locally still WIP
Release schedule
—> Everything under 0.11 will be backward compatible (except for a bug that will be fixed)
-> epoch-basis instead of slot basis for (?)
-> Release today
—> 0.12 for IETF draft 7 of the hash-to-curve
-> will include non-substantive type fixes
-> additional validation will be integrated in the fork choice
0.12 not interoperable with 0.11
-> maybe with fork choice tests
-> might come with modified rewards & penalties tests (bug in rewards & penalties in Prysm/Lighthouse)
-> function will probably be rewritten and split so that tests are with more granularity
-> proposal in 24 hours
-> Topaz testnet: will maintain a fork for interop
Proto - The rewards/penalties: doesn’t affect users, for different types of (?) you can’t get slightly different rewards. —> Rewards are slightly higher than they should be.
nimbus, pyspec, zcli also impact
Network testing
Rumour updated for new RPC types, clients can start testing with this now.
For testing 7+ clients instead of instrumenting all clients will take an approach more similar to the current consensus tests
Beacon Fuzz
Found 5 bugs on Nimbus including stack overflow
Found 2 bugs 2 bugs on zrnt
Found SSZ deserialisation bug in Lighthouse
Have been using HongFuzz
—> Risky to treat beaconstate as a trusted structure
especially if we use warp sync in the future
Client updates
Spec compliance, passing 0.11 tests
Gaps are Snappy compression and ENR
Topaz testnet, 121k validators
thousands (?) epochs so far since Friday
didn’t skip finality except 1
Can sync with lighthouse until epoch 1 (consensus bug in Prysm)
Great feedback, focus on (?) and UX
JVM libp2p Noise interoperable with lighthouse
Protoarray fork choice implemented
Hot state management rework to keep fully in memory
Focus: finishing network changes.
Steady progress on v0.11 testnet
internal changes to the fork choice
Update BLS to
Comment by Paul on update/rewind
- Skip signature verification when applying blocks (for blocks from cold storage)
saves a significant amount of time
Launched new testnet
Master is now 0.11.1 state transition
Almost fully compatible (networking spec to do)
Learning Ansible to automate the testnet (
sync with Prysm
RPC working
Rust new futures “stable futures”
Wanted to implement safe-arithmetic to have the implementation panic-free.
Afr Schoe
—> 0.11
Research updates
- monitoring tool for Topaz
- JIT problems of EE
- Transpiler for pyspecs to test fork choice
- Disc v5 research
LMD votes of attestations not being consistent
Fork choice testing should be in next updates of spec
Orchestrated attack
- Eth1x, nothing new since last announcement
Getting more data to evaluate how to partition Eth1 transactions if split across shards
- Writeup coming
-> Time to stabilise phase 1 spec
API call
Marin: small API combined
Danny’s goal: have them for public review by next week
Spec discussion
Vitalik: discussion on phase 1
Open discussion
Preston: release schedule
0.11.2 today (backward compatible)
0.12.x pending hash to curve (imminent)
Afterwards: critical security fixes
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment