All Core Devs - Testing (ACDT) #88, July 20, 2026

Protocol Calls & happenings
,
1

Agenda

Devnet status updates

  • Updates on glamsterdam-devnet-7

Discussions

EL breakout

CL breakout

TBD

Meeting Time: Monday, July 20, 2026 at 14:00 UTC (60 minutes)

GitHub Issue

2

Video, transcript & chatlog

News coverage

Resources

3

Meeting Summary:

The R4Dev Testing 88 meeting focused on updates and discussions regarding Glamsterdam DevNet 7 and upcoming DevNet 8. Stefan reported that DevNet 7 was performing well with most clients onboarded, though Aragon and Grandin would have fewer validators initially. The team discussed adding more testing scenarios and chaos testing to DevNet 7, with plans to schedule DevNet 8 for the following week or the week after. Clients reported some performance issues with stable containers being 10-20% slower, and there were discussions about JSON RPC inconsistencies across different clients that needed alignment before DevNet 8. The team also addressed fork transition bugs, particularly from Lodestar regarding stack overflow issues with the spread operator, and discussed the status of EIPs ready for SFI stage promotion.

Click to expand detailed summary

Justin Traglia opened the R4Dev’s Testing 88 meeting and welcomed attendees. He mentioned an agenda item related to the Ethereum project and asked Stefan for updates on Glamsterdam DevNet 7.

Stefan reported that Glamsterdam DevNet7 is progressing well with most clients onboarded, though Aragon and Grandin will have fewer validators initially. He plans to add more testing scenarios to DevNet7 and work on specifying DevNet8, including Maria’s requested repricing EIP changes and additional fork transition tests. Justin inquired about missing slots for PRISM and PRISM-Aragon on DORA, which Stefan attributed to clients still syncing, and Justin asked about timeline expectations for DevNet8 but the response was cut off.

Stefan suggested postponing DevNet 8 to next week or the week after, as there is still significant testing ground available with DevNet 7 and many people are on holiday. Parithosh mentioned updates to a malicious PRISM node and noted that similar testing scenarios like Death Star would be available this week. The team discussed plans for future DevNets, with Parithosh proposing a smooth transition in DevNet 8 followed by a non-finality-focused DevNet 9 in the coming month.

The team discussed issues with BAL specific JSON RPCs across clients and agreed on the need for alignment. Cayman explained that Lodestar experienced a stack overflow bug due to the spread operator being used with too many arguments when handling large Merkle trees. The discussion highlighted the need for more comprehensive testing to prevent future differences in RPC implementations.

The team discussed progress on merging devnet-7 to master and ongoing work on the builder API implementation. Matthew suggested reaching out to application-level stakeholders like Cowswap and builders to practice the fork transition on devnet-8. Justin inquired about notable issues from EL clients and benchmarking results for SSD changes, with Saulius reporting that Grandini found significant performance slowdowns in some areas that would need optimization.

The team discussed performance issues with stable containers, noting they were 10-20% slower compared to non-progressive containers. Enrico mentioned they would need to conduct further optimization rounds to better understand data structure behavior. Stefan reported that DevNet 7 had mainly gas accounting inconsistencies, with the biggest issue being SSC containers during fork transitions due to validator set size rather than exited validators.

The team discussed making DevNet 8 use disk V5 only, with Csaba planning to test this change before implementation. Parithosh confirmed that QUIC is enabled in DevNet 7 for sales purposes. Pooja inquired about upgrading the stage to SFI in DevNet 7, noting that all four SFI criteria from EIP 7723 appear to be satisfied, though no one could provide an update on the status of this upgrade.

The team discussed moving several proposals to the SFI stage, with Parithosh confirming that the PandaOps team was ready to SFI the proposals from a testing perspective on ACDT. Pooja proposed moving EIP 7773 Meta E4 Glamsterdam upgrade to review status, which received no objections from the team. The group agreed to proceed with moving all proposals to SFI stage pending ratification in the upcoming ACD call on Thursday, with authors needing to add their PRs or approvals to the EIP editing office hour for consideration before the next ACD call.

Toni discussed JSON RPC implementation differences between clients and requested teams review the summarized points in a HMD document. Stefan mentioned that Chase has been working on JSON-RPC tests and offered to connect with him about the implementation. The discussion focused on ensuring consistent responses across clients, particularly regarding error codes and empty list handling, with Toni noting a recent PR against execution APIs that addressed these divergences.

The team discussed adding gas state information to subcode traces, with Dragan raising concerns about implementation details while Toni confirmed this was independent from Baal-specific methods. The group agreed to proceed with the tracing changes and align on a PR for the EIP. Toni also presented a PR for EAP7997 related to contracts factory, which adds the factory to the EOS via an irregular state transition to align with GEF’s approach of checking and writing contract addresses when the factory is not pre-deployed.

Mario explained the context of an issue raised by Ignacio regarding checking contract existence that affects execution witnesses, noting that a PR was opened to remove this check to avoid inconsistent behavior between clients. Toni discussed finding differences between Aragon and Geth JSON RPC testing related to fork boundaries and factory deployment. Mario shared a link to the relevant PR that removes state transition mentions and explains how to handle cases where the contract is not present, relying on gas schedule limitations and irregular state transitions.

Marius expressed that a specific specification discussed in the team was not necessary for mainnet and there was no strong argument for implementing it. Toni agreed and decided to close the PR since it was not critical. The team ended the conversation early as there were no additional topics to discuss.

Next Steps:

  • Stefan: Onboard Aragon to Glamsterdam DevNet 7.
  • Stefan: Add more testing and chaos scenarios to DevNet 7.
  • Stefan: Spec DevNet 8, including minor cleaning up changes to repricing EIPs and plan for fork transition tests.
  • Parithosh: Review and merge codebase for malicious PRISM node after Aussies review it.
  • Toni: Clients should review the Book Club Access List Discord for BAL-specific JSON RPC differences and align them for DevNet 8.
  • Stefan: Reach out to Chase regarding JSON RPC tests and discuss adding Glamsterdam tests.
  • Csaba: Implement and test code to ensure DevNet 8 runs with disk V5 only.
  • Parithosh: Verify if QUIC is enabled in DevNet 7.
  • Pooja: Authors should add PRs for EIP 7773 Meta E4 Glamsterdam upgrade to the EIP editing office hour for review and approval before the next ACD call on Thursday.
  • Toni: Close the PR related to EAP7997 (contracts factory) as it was decided not to include it in EELS.

Recording Access:

4

YouTube recording available: https://youtu.be/LIi4RY-TrdU

5

CL breakout YouTube recording available: https://youtu.be/fe0Q3AbHwd8