All Core Devs - Execution (ACDE) #199, October 24 2024

abcoathup · October 12, 2024, 12:11am

timbeiko · October 24, 2024, 5:39pm

Action Items

EIP-7742 is Scheduled for Inclusion in Pectra, teams to review @g11in’s proposed change.
BLS MSM Repricings: @chfast will propose a new lookup table for costs and EL teams should run benchmarks with these. Further discussion planned on Monday’s testing call.
EIP-7685 Empty Requests: EL + CL teams to review @fjl’s PRs by next ACDC, ideally merge before the call.
EIP-7702 & EXTCODE* : @frangio to update his PR to return a constant value rather than the delegation designator when calling EXTCODE*.
eth_* API Revert Error Codes: teams to review @fjl’s proposal to standardize revert error codes.

DST will start in some regions prior to the next ACDE, check your local time.
The last pre-devcon ACD is ACDE#200 (!!) and first post-devcon ACD will be ACDC#146. More explicitly we will:
- Cancel 11/14 ACDC
- Cancel 11/11 and 11/18 Pectra Interop Testing Calls
- Replace 11/21 ACDE by a Pectra Interop Testing Call

Good participation rate on the network
Issues with Erigon, EthereumJS (investigating), and Grandine (WIP fix)
Bad block generation has started, currently running on 1 node, plan to increase to 4 soon

Keep subgroup checks: community outreach revealed that many STARK-based projects are unlikely to switch to BLS12-381 due to the lack of infrastructure support, and no project expressed a strong desire to have the non-subgroup-checked versions of the precompiles. Additionally, further investigation into the BLS libraries showed that it would require non-trivial changes there to separate subgroup checks out. We’ll keep the subgroup checks in the precompiles.
MSM Repricings: there was no consensus on the MSM repricing values. @chfast plans to propose a new lookup table for costs and EL teams will run benchmarks based on these. More conversations to follow in next week’s testing call.

@fjl proposed excluding empty items from requests_hash commitments (PR), to keep the empty requests independent of the fork schedule, simplifying genesis configurations in clients.
EL clients supported the change, with some tentative support from CL teams. It was added to the devnet-5 spec. If there are still issues by next week’s ACDC, we’ll discuss them there.

@frangio opened a PR to address delegation concerns related to EXTCODE* raised on EthMagicians
There was some support for the change on the call, but @adietrichs proposed that instead of operating on the delegation designator itself, the calls to EXTCODE* return a constant value.
@frangio will open a PR incorporating the changes

EL teams agreed to move the EIP to SFI (PR.
@g11in proposed a change to the EIP, but due to mic issues & call attendees not having properly reviewed, we’ll review async

Note: this was a contentious topic with a lot of back and forth, please watch the livestream for the full nuance.

@Giulio2002 proposed EIP-7790 to suggest values to use as part of the EIP-7783 gradual gas limit proposal: going from 30m → 60m gas over a 2 year period.
Since the last call, Giulio opened PRs for EIP-7783 on Reth, Erigon and Geth, while the Nethermind team has begun work on their own implementation.
@Giulio2002 shared that he’d reached out to client teams to get their feedback on the 7790 numbers. He said that Nethermind and Erigon were onboard, @karalabe had expressed some support (although others in Geth opposed on the call) and Besu had concerns.
On the call, several concerns were raised about the proposal, including:
- The risks of raising the gas limit before implementing EIP-7623
- Its impact on history growth (although somewhat mitigated post-4844)
- Potential integration concerns with mev-boost
- Concerns that a gradual raise may lead to “silent failures”, as opposed to our current approach of “step function” increases
- Questions about whether 60M over 2 years was the right number of aim for
- Concerns about pre-committing to long-term increases
While there is value in client teams being aligned on their approaches here, it is worth noting that the gas limit is not controlled by clients/hard forks, but by validators when they propose a block. This means that client teams can suggest different defaults, which block proposers may or may not adhere to.

@fjl proposed adding a standard error code for reverts in eth_* APIs (PR), mirroring Geth’s behavior since 2022.
There was broad support, conversation to continue on the PR