All Core Devs - Testing (ACDT) #79, May 11, 2026

Protocol Calls & happenings
,
1

Agenda

Agenda

Agenda

  • Summary of interop event
  • glamsterdam devnet updates
  • spec glamsterdam next devnet
  • debug RPC update

Meeting Time: Monday, May 11, 2026 at 14:00 UTC (60 minutes)

GitHub Issue

2

Video, transcript & chatlog

News coverage

Resources

4

Meeting Summary:

The meeting focused on reviewing the interop event updates and discussing various technical topics related to the execution layer and consensus layer integration. The team reviewed progress on Glamsterdam DevNet, block-level access lists, repricing optimizations, and Fossil DevNet testing. A significant portion of the discussion centered on gas limit handling, where the team agreed to shift responsibility for gas limit validation from the execution layer to the consensus layer, with the consensus layer validating against proposal preferences while the execution layer would handle the EIP-1559 rule calculations. The group also discussed a PR regarding execution API changes and decided to close an existing PR for SSE limit increases. Additional topics included benchmarking for signature verification costs and potential attack implications of raising the min builder withdrawability delay.

Click to expand detailed summary

The team discussed updates from the interop event, focusing on various workstreams including Glamsterdam, block-level access lists, and repricing. Barnabas reported that Glamsterdam started well but faced stability issues, particularly with clients unable to sync on Lighthouse and envelope serving problems from Teku. Stefan shared progress on block-level access lists, noting that EIP-4895 is near finalization and plans to incorporate benchmarking into EIP-4897 while merging everything into Amsterdam DevNets. Maria provided an update on repricing optimizations, mentioning preliminary numbers for EIP-7904 and EIP-8038, with plans to provide more concrete updates next week after further optimization work.

The team discussed issues with Glamsterdam DevNet3, where Potuz identified problems with peers serving inconsistent data, particularly regarding block payloads. Enrico suggested investigating logs to understand the sync issues better, noting that similar problems exist with other clients like Lodestar. The team agreed to continue troubleshooting in the EPBS channel, with Barnabas coordinating potential recovery efforts or launch of DevNet4 if needed. Parithosh also mentioned a PR to raise the min builder withdrawability delay to 8192 epochs, which aims to make it more expensive for attackers to spam new builder deposits, with no objections raised during the discussion.

The team discussed concerns about signature verification performance and potential attack vectors for builder deposits. Justin proposed holding off on implementing a queue-based approach until more testing is conducted on extending the withdrawal delay as a potential solution. Pawan shared benchmark results showing that verifying 8192 signatures took around 700 milliseconds on an M1 MacBook, with performance degrading slightly with invalid signatures. The team agreed to further validate these benchmarks and quantify the actual costs of potential attacks before making decisions. They also briefly discussed a PR related to increasing the SSE limit, with the consensus to close it for now rather than making immediate changes.

The team discussed changes to how gas limits are handled in the consensus and execution layers. They agreed that the CL should handle target gas limit preferences and validation against these targets, while the EL would only need to enforce the 1/1024 rule for gas limit changes. Nico was tasked with updating the engine API PR, and Bharath will create a consensus spec PR to reflect these changes. Barnabas suggested a separate EIP for gas limit scheduling via BPO, which will be discussed further in future meetings. The team decided to address DevNet scoping asynchronously or in the next ACDC meeting.

Next Steps:

  • Barnabas: Coordinate recovery of Glamsterdam DevNet 3, or if unrecoverable, launch DevNet 4
  • Enrico/Teku team: Investigate logs to identify why Teku is serving inconsistent block/payload ranges that cause peers to be downscored, and fix the bug related to self-built bids
  • Potuz/Prism team: Continue investigating sync issues on Glamsterdam DevNet 3 and share findings in the EPBS channel
  • Justin: Continue benchmarking the cost of spammed new builder deposits (signature verification performance), validate numbers, and quantify attack costs (capital required, interest cost, timeline) to present at ACDC
  • pawan: Complete and validate the Lighthouse benchmark for 8192 builder deposit signature verifications and post results to the channel
  • Justin: Close consensus specs PR #5224 (raising SSE limit), with the note to reopen if stable containers are not included in Glamsterdam
  • Stefan: Share the BAL DevNet 7 spec as a gist (due to HackMD issues) and scope BAL DevNet 7 incorporating benchmarking from DevNet 3 and making EIP-70/71 required
  • Maria: Finalize updated EIP numbers for 7904 (Compute EIP) and 8038 (State Access EIP) with stable optimization numbers; provide concrete update on 7904 by next week
  • Toni: Merge the Execution API PR (replacing error code, adding block access list hash to header, adding debug.getRawBlockAccessList method) after the call
  • Nico: Update the existing engine API PR to add gas limit target to payload attributes (replacing the EL-side gas limit flag with CL-provided target), for discussion at ACDC this week and merge by next week
  • Bharath: Raise a consensus spec PR to update the gas limit check on the bid/envelope (changing from strict equality to target-based validation per the agreed approach)
  • Barnabas: Write up an EIP proposal for making gas limit a BPO-configurable parameter and bring it up at ACDE
  • Parithosh/All: Conduct BAL DevNet 7 and Glamsterdam DevNet scoping async, or address it first at ACDC this week

Recording Access:

5

YouTube Stream Links: