All Core Devs - Testing (ACDT) #81, June 1, 2026

Protocol Calls & happenings
,
1

Agenda

Glamsterdam

Heads up: As repricing specs have stabilized, the fortnightly Wednesday Gas Repricing Breakout Calls held by @misilva73 have stopped; any required discussion will happen in ACDT.

Devnet Status and Benchmarking

  • bal-devnet-7, hive dashboard, dora:

    • Stable, 100% EVM fuzzing participation (reported in Repricings #8) @qu0b.
    • Clients:
      • eth/70 (EIP-7975) and eth/71 (EIP-8159) are mandatory on bal-devnet-7 - is this work complete? @qu0b.
    • Final test release tests-bal@v7.3.0 once 8037 lands in forks/amsterdam, tracker: ethereum/execution-specs#2912 @spencer-tb.

  • Benchmarking / repricing status (@misilva73 + client teams):

    • State of client BAL optimizations in bal-devnet-7 branches? Per-client round.
      • ethrex reports ~3x exec speed-up vs devnet-3?
      • Nethermind OOM affecting benchmarking results?
      • Besu (shipped AOT and an OOM fix).
      • Geth BAL opt branch this week?
    • How much of a priority is it to replace the perf-devnet-3 snapshots with ethereum/state-actor? (execution-specs#2916). This can be deferred to tomorrow’s Gas Lighting call.

  • glamsterdam-devnet-4 not so healthy as of ~May 25::

    • Chain split ~slot 22800; prysm-nethermind-1 forked on EMPTY after missing the 22800 payload.
    • PR (consensus-specs#5307).
    • Recover the chain or abandon for devnet-5?
Spec Changes
Glamsterdam spec next TODOs
  1. Finalize gas repricings numbers (ongoing on bal-devnet-7).
  2. CL stability; resolve consensus-specs#5307.
  3. Include new EL EIPS:
    i. EIP-8246 (remove self-destruct burn).
    ii. EIP-2780 (intrinsic gas) and EIP-8038 (state-access gas cost update), requires 1., but not 4.
  4. Updated EL EIP: EIP-7954 (increase max contract size w/larger numbers).
  5. EL EIP spec changes as above.
  6. Any other CL changes?

i. EIP-8045: Exclude slashed validators from proposing.
ii. EIP-7688: Forward compatible consensus data structures @etan-status?
iii. …

Options and scoping of glamsterdam-devnet-5 OR next devnets
  1. Clean glamsterdam-devnet-4 spec relaunch. (CL ask from ACDC #179: stability first, no scope increase for 2 weeks).
  2. glamsterdam-devnet-4 with bal-devnet-7 (with or without some SFId EL EIP spec changes, details below).
  3. As 2. but with bal-devnet-7 with 2780/8038 placeholder numbers.
  4. Launch two devnets simultaneously and merge into later devnet:
    i. glamsterdam-devnet-5 for CL stability.
    ii. bal-devnet-8 for close-to-final EL spec.
    iii. → merge into glamsterdam-devnet-6.
  • Timing of devnet launch(es)?
Other Topics

Please comment on this issue with other topics!

Last call:

Meeting Time: Monday, June 01, 2026 at 14:00 UTC (60 minutes)

GitHub Issue

1 Like
2

Video, transcript & chatlog

News coverage

Resources

3

Meeting Summary:

The meeting focused on testing updates for Ethereum core development, specifically discussing the current state of Bell DevNet 7 and the aftermath of DevNet 4. Stefan reported that EIP-8037 implementation was mostly successful across clients with only minor blips during updates, and most clients were now working on benchmarking performance. The team addressed a recent chain split in DevNet 4, with Barnabas explaining that a clarification was needed regarding block building and acceptance, and the group decided to include EIP-8045 in the next release after receiving support from multiple client teams including Lodestar, Prism, and Lighthouse. The discussion then moved to several EIP specifications, including a new requirement for at least one storage change to trigger a slot change, and Tony’s proposal for state access decoupling which would require implementation changes from most clients except Ref. The team also reviewed EIP-8037 changes regarding authorization state gas calculations and EIP-2780 modifications for ETH transfer pricing, with Maria noting that benchmarking results were still needed before finalizing these changes. The conversation ended with extensive discussion about the structure of future DevNets, with debate over whether to launch Glamsterdam DevNet 5 as a CL-only update focused on stability or to include additional EL changes, ultimately deciding to continue this discussion asynchronously due to concerns about rushing optimization work.

Click to expand detailed summary

The meeting focused on the current state of DevNet 7, which was reported to be stable with 100% EVM fuzzing coverage. The team discussed that the fortnightly Wednesday gas repricing breakout calls led by Maria have been discontinued as the specs have stabilized, with discussions now moving to the ACDT calls. The meeting also covered upcoming SPAC updates and plans for DevNet 5, though specific details about these were not provided in the transcript.

The team confirmed that both EVM fuzzing updates (EVE 70 and EVE 71) have been successfully implemented across all clients including Nevermind, Besu, EthRex, Aragon, Nimbus, and Geth. Stefan reported that the implementation is stable with most clients now focusing on benchmarking and performance testing. The team plans to cut a final release for Bell DevNet 7 at 7.3.0 once Steel completes merging E8037 into their default development branch for Amsterdam.

The team discussed updates on Bell DevNet 7 and Glamsterdam DevNet 4. For Bell DevNet 7, Maria suggested deferring updates until the next day’s call due to ongoing client optimization work. Regarding Glamsterdam DevNet 4, Barnabas explained that a chain split occurred last week, which has been addressed through updated specifications. Justin suggested merging several PRs before creating the Alpha 9 release, including fixes for should_build_on_full and other related issues.

The team discussed including EIP-8045 in the next devnet release, with multiple client teams including Lighthouse, Nimbus, and others confirming they have implementations ready. Barnabas noted that EIP-8045 is a simple change and most teams support including it. The discussion then moved to spec changes, specifically mentioning EIP-7928 which requires slot changes to have at least one storage change.

The team discussed changes to gas limit handling, confirming that clients will need to make changes when moving to the Glamsterdam series of DevNets. Rahul explained that a new protection mechanism against malicious files with bloated entries cannot be generated by clients and is intended to prevent specific security issues. Toni presented a PR for EIP-7928, which proposes to decouple state-access validation from EIP-2929 by reordering gas charges, with Dragan explaining that this approach would maintain performance optimizations in the current implementation while addressing concerns about check ordering.

Toni discussed a potential PR merge that would require most clients to change their implementation regarding bulk access handling, noting it would be a significant code change. Dragan confirmed their team is satisfied with the current implementation and would either accept it as-is or with the latest PR. The team agreed to have clients review the complexity of the changes by ACDE on Thursday, with Gary committing to check it. Maria then highlighted two open changes to EIP-8037 that were meant for Glamsterdam DevNet 5 but will likely move to DevNet 6, including a merged PR for calldata floor accounting and an open PR by Gary regarding authorization gas refunds.

The team discussed a proposed change to authorization list state gas calculations, with Maria explaining that while she supports the change, it would require implementation adjustments. Dragan and Spencer provided positive feedback on the proposal, though Dragan noted it would need detailed review. The meeting then shifted to EIP-2780, where Maria explained that a proposed simplification to self-transfer and precompile charging was on hold pending refactoring and benchmarking results, with the team agreeing to revisit this topic the following week.

The team discussed updates to the GLAMsterdam DevNet, focusing on CL stability and EL specifications. They agreed to prioritize optimizations and the self-destruct burn removal (EIP-8246) over other changes like factory precompile (EIP-7997). The group decided to keep GLAMsterdam DevNet 5 as a CL-only update, with plans to potentially include EL changes in GLAMsterdam DevNet 6. They also discussed the possibility of launching two devnets simultaneously but ultimately decided to postpone that decision. The team agreed to continue the discussion asynchronously to finalize the details for the next devnet releases.

Next Steps:

  • Stefan: Finalize benchmarking on ValDevNet 7 and collect final performance numbers from all clients.
  • danceratopz: Cut a final release for Bell DevNet 7 (v7.3.0) once Steel merges AT37 into their default development branch, and ping Discord when ready.
  • Maria: Defer gas benchmarking optimization updates to the Gas Benchmarking Committee call tomorrow and provide updates when final numbers are available.
  • Barnabas: Prepare Alpha 9 spec release (including AT45) targeting end of day, with aim to launch Glamsterdam DevNet 5 as early as tomorrow.
  • Justin: Review and merge the open PRs (including AT45-related ones) before the Alpha 9 spec release, targeting today or early tomorrow.
  • Gary: Review the EIP-7928 (BAL state access decoupling) PR by ACDE on Thursday and provide feedback on implementation complexity.
  • Toni: Add EIP-7928 (BAL state access decoupling) to the ACDE Thursday agenda for a follow-up decision.
  • Maria: Finalize parameter changes PR for EIP-2780, then align with Dragan’s simplification PR afterward.
  • Dragan: Update the EIP-2780 simplification PR or coordinate with Maria to make changes on top of her finalized PR.
  • All EL client teams (Aragon, Nimbus EL): Review Gary’s EIP-8037 PR on reservoir mechanics for authorizations and provide feedback before the next DevNet.
  • danceratopz: Open a thread in ETH R&D after the call to outline the plan for Glamsterdam DevNet 5 (CL-only changes with AT45), and invite feedback especially regarding inclusion of EIP-8246.
  • Carson: Complete review of Steel Team’s implementation and tests for EIP-8246 (self-destruct burn removal) today, targeting inclusion in the next EL DevNet after DevNet 5.
  • All client teams: Continue benchmarking optimizations on ValDevNet 7 as the current priority before focusing on new EIPs.
  • danceratopz: Bring the decision on EIP-7997 (deterministic factory precompile) and EIP-7928 back to ACDE next week, along with the decision on whether the next EL DevNet will be Glamsterdam DevNet 6 or ValDevNet 8.

Recording Access:

4

YouTube Stream Links: