All Core Devs - Testing (ACDT) #47 | August 4 2025

Protocol Calls & happenings
,
4

ACD Testing Call #47 - (Quick Notes)

Facilitator: Mario Vega

TL;DR

  • Fusaka Devnet 3
    • Nimbus and Lodestar are still being debugged; fixes are underway.
    • Decision: Wait until both are stable before advancing.
  • EIP-7825 Gas Limit
    • Proposal to raise to 30M discussed; concern over app impact.
    • Decision: Maintain current 16M gas limit on Devnet 3.
  • Gas Limit Testing
    • Stateful and large contract tests in progress; repo flag awaited.
    • ModExp pricing PoC tool being developed; Geth change not expected soon.
  • EIP-7951 (secp256r1) Benchmarking
    • Besu results show high cost.
    • Action: Mario to follow up with Besu team asynchronously.
  • INTERVALS_PER_SLOT Replacement (EIP-4476)
    • Adds slot timing config without protocol changes.
    • Decision: No breaking changes; clients should begin prepping for implementation.
  • PR-4477 Proposal
    • Dev’s consensus that it’s too late for Fusaka.
    • Decision: Do not proceed; no Devnet 4 for this. Plan shadow fork next.
  • Sunnyside Devnet / PeerDAS Syncing
    • Sync failures in perfect PeerDAS; investigation ongoing.
    • Decision: Feature required, clients to report progress.
  • Backfill Implementation
    • Prysm, Lighthouse, and Lodestar in progress.
    • Decision: Next ACD to revisit Devnet 4 for backfill.

Fusaka Devnet 3

Barnabas:

  • Devnet 3 is mostly stable with ~90% participation.
  • All 5 Blob Parameter Only (BPO) forks are activated and running smoothly.
  • Most prior issues have been resolved.
  • However, 2 consensus clients — Nimbus and Lodestar — are still facing issues.
  • Caplin also involved in ongoing testing.

Lodestar:

  • Matthew K is investigating genesis value sync and working on resolving the bug.

Nimbus:

  • Update relayed by Agnish:
  • Issue with the custody column related to hash functions in indices.
  • EF test vectors are passing.
  • EAS (Earliest Available Slot) was not moving — now fixed.
  • General updates: MEV issues were reported and have been addressed.
  • Issue raised on Eth R&D Discord - Nimbus and Geth having difficulty syncing (likely a peering issue).
  • DAS Guardian fix expected within the next few days.

Bharath:

  • Local transaction propagation was disabled on the RETH node, which caused transactions not to be streamed.
  • Fixes of issues shared on the call.
  • Also, under discussion with the RETH team to consider removing local transaction propagation entirely.
  • Some complications related to spam filtering reported, and the root cause of transaction exclusion has now been identified.

Roman (RETH): Inquired about the RETH flag and its specific issue.

Bharath: Shared technical context also shared on Eth R&D Discord.

Nethermind (Marek):

  • Nethermind worked with rbuilder as state provider on integration with state
  • The state sync component is ready, but the EVM integration is still in progress.
  • Additional testing is requested before full readiness.
  • More details: Flashbots Multi-client Support Announcement

Conclusion & Next Steps

Mario:

  • Overall, Devnet 3 is progressing well.

Barnabas:

  • Aiming to ensure Nimbus and Lodestar are bug-free before moving forward.

Raising the Limit of EIP-7825 to 30 Million

Mario:

  • Provided an update from last Thursday’s ACD discussion.
  • Discussion link on Ethereum Magicians.
  • Raised concerns that raising the gas limit may negatively impact certain applications.

Toni:

  • Stated that tradeoffs favor keeping the current spec as implemented in Devnet 3.

Decision:

  • Maintain the gas limit at 16 million for now.

Gas Limit Testing Updates

Barnabas:

  • No major updates.

Kamil:

  • Still working toward full state testing.
  • Marcin is running additional tests and benchmarking.
  • EEST is running on existing infrastructure for stateful testing.
  • A test is live on a contract with a very large state size; waiting on repo to receive the necessary flag.
  • Some tweaks are needed on Devnet.
  • BloatNet is being prepared; results expected this week.
  • Will continue iterating and updating.

Marcin:

  • Developing a new tool to evaluate ModExp pricing.
  • Differences in gas usage are expected; further investigation underway.
  • Currently working on a PoC — actual testing still pending.

Parithosh (in chat):

  • Geth team noted that upstreaming changes would take time, if at all. Not a short-term solution.

Mario:

  • No Geth representative was present to elaborate further.

Louis:

  • Related issue: execution-spec-tests#1976
  • Related PR for configuring gas limits: PR #1983

Kamil:

  • 60M gas limit update still requires more testing.
  • Genesis file handling and stateful test considerations are ongoing.

Besu’s EIP-7951 Benchmarking Results

Justin (Besu): Updates shared in Comment #1648.

Mario:

  • Initial results suggest the cost is too high. Will follow up asynchronously with the Besu dev team.

Replace INTERVALS_PER_SLOT with Explicit Slot Component Times

Justin Traglia:

  • Introduced the PR to replace INTERVALS_PER_SLOT with explicit slot component times.
  • Goal: Raise awareness now and propose merging within the next 3 months to begin implementation.
  • This proposal adds additional config values; no substantial changes to protocol behavior.
  • Believes there’s no reason to delay merging sooner.

Mario:

  • Asked whether this introduces any breaking changes for Fusaka.

Justin T.:

  • No breaking changes expected.
  • Clients will need to update configurations, but should not impact Fusaka.

Pawan:

  • Initial review looks good.
  • Will revisit the PR for a deeper look.

PR-4477 Discussion & Decision

  • Barnabas shared and requested decison on PR.

Feedback from devs on call & chat:

  • Justin T.: Against including this change.
  • Alex S.: Suggests leaving it for now.
  • Pawan: Opposes the proposal.
  • Barnabas: Notes that the proposal has received support from some CL devs (Potuz, Terence, nflaig).
  • Toni W.: Feels it’s already too late to consider this change.
  • Csaba K.: Also against; if needed, should be done with a future BPO.
  • Barnabas: Opposes bundling this change with BPO.
  • Mario: Including this would require an entirely new fork.
  • Pari: Suggests deferring it to Glamsterdam.
  • Raul: Reminds that BPOs were explicitly designed to avoid code changes.
  • Agnish: Opposes inclusion; the proposal is too tricky and more data is needed from dynamic DA conditions with supernodes.
  • nflaig (Nico): Agrees it’s too late for Fusaka.

Decision & Next step

  • This PR will not move forward.
  • Devnet 4 will not be launched.
  • Focus will remain on stabilizing Nimbus and Lodestar before further progression.
  • A shadow fork is being planned for next week. Barnabas encourages client teams to have established branches ready ahead of the shadow fork.

Sunnyside Update

Update on New Report:

  • Successfully reached 60 blobs per block; observed impact on usage.
  • Hit limits as expected per hardware requirements in EIP.
  • Encountered genesis sync failures in perfect PeerDAS network setup.

Devs comments (from call & chat):

Manu:

  • Provided explanation for Prysm issues on perfect PeerDAS.
  • Actively working on a fix; expects resolution by end of the week.

Agnish:

  • Requested clarity on the importance of syncing in perfect PeerDAS mode.

Pari:

  • Clarified that Fusaka will not support PeerDAS with this in its initial release.
  • This feature can be evaluated later, and if necessary, included in a future BPO.

Agnish agreed with this direction.

Mario:

  • Agreed Perfect PeerDAS support is not required for Fusaka.
  • May be revisited for future upgrades.

Agnish:

  • Shared that reconstruction-based syncing offers more guarantees and would prefer that over relying on dynamic peer discovery.
  • May need this feature.

Mario:

  • Acknowledged that this feature might become necessary.
  • Requested CL teams to confirm timelines for implementation as it may be a potential blocker for Fusaka.

Pawan:

  • Currently does not have it implemented but is aware of the mechanism and will assess how long it would take to develop.

Decision:

Mario:

  • This is important enough to be tracked closely.
  • Requested the rest of the CL team to respond.
  • Will raise the topic again on the next ACD call for broader CL input.

Note:
Sunnyside may spin up additional testnets to support further PeerDAS testing.

Backfill

Barnabas:

  • Asked if any client implementations for backfill are in progress.

Manu:

  • Prysm implementation is in progress.

Other Clients:

  • Lighthouse and Lodestar are actively working on it.

Devnet 4 for Backfill?

Barnabas:

  • Proposed the idea of a Devnet 4 focused on backfill testing.
  • Suggested waiting for confirmation on ACD once client implementations are ready.

Backfill Implementation Challenges

Agnish:

  • CPU usage is significant, depending on how quickly backfilling is expected.

Question raised: Should nodes provide slot-level data granularity while backfill is in progress?

Pawan:

  • Mentioned they had to change the database structure to make Earliest Available Slot (EAS) accessible.
  • Average backfill size observed: ~1 epoch.
  • For their implementation, this is not a blocker.

Conclusion:

  • Awaiting updates on perfect PeerDAS syncing before proceeding further.

Suggest any changes to notes here.

1 Like