AllCoreDevs - Testing #081
Transcript
- danceratopz
📋 GM! ACDT #81 agenda <https://github.com/ethereum/pm/issues/2085>
- Louis
GM!
- Justin Traglia
Gm
- danceratopz
📋 GM! ACDT #81 agenda https://github.com/ethereum/pm/issues/2085
- danceratopz
🟢 bal-devnet-7 - stable, 100% EVM fuzzing hive <https://hive.ethpandaops.io/#/group/bal> dora <https://dora.bal-devnet-7.ethpandaops.io/> 🧪 final test release tests-bal@v7.3.0 once 8037 lands in forks/amsterdam tracker <https://github.com/ethereum/execution-specs/issues/2912>
- danceratopz
Replying to "🟢 bal-devnet-7 - st..." eth/70 (EIP-7975) and eth/71 (EIP-8159) are mandatory on bal-devnet-7 - is this work complete?
- Toni Wahrstätter
eth/71 is done in all clients afaik
- Daniel Lehrner (Besu)
Both implemented in Besu
- Edgar | ethrex
same in ethrex
- Dragan Rakita
Just landed
- danceratopz
🏎️ Benchmarking - state of client BAL optimizations on bal-devnet-7 branches? (per-client round) state-actor (replace perf-devnet-3 snapshots) <https://github.com/ethereum/state-actor> issue <https://github.com/ethereum/execution-specs/issues/2916>
- danceratopz
---------- Devnet Status: glamsterdam-devnet-4 post-mortem ---------- 🔴 Status: <https://dora.glamsterdam-devnet-4.ethpandaops.io/> 🪦 glamsterdam-devnet-4 - recover or abandon? TLDR: ePBS split ~slot 22800; defect open (peering vs topic-ID vs should_build_on_full / PTC timing) <https://github.com/ethereum/consensus-specs/issues/5307>
- Justin Traglia
What did we change in the spec?
- Eitan Seri-Levi
the should build on full thing
- Nico Flaig
- Justin Traglia
I would like for these to be merged before making the alpha.9 release btw: * https://github.com/ethereum/consensus-specs/pull/5306: Apply proposer boost if dependent roots match * https://github.com/ethereum/consensus-specs/pull/5309: Limit should_build_on_full checks to the previous slot * https://github.com/ethereum/consensus-specs/pull/5310: Use should_build_on_full for bid parent block hash * https://github.com/ethereum/consensus-specs/pull/5174: Remove should_override_forkchoice_update And if we decide to include it: * https://github.com/ethereum/consensus-specs/pull/5115: Exclude slashed validators from proposing (EIP-8045)
- Eitan Seri-Levi
We had a bug where our node couldn’t restart under certain conditions
- Nico Flaig
what about https://github.com/ethereum/consensus-specs/pull/5210 ?
- Justin Traglia
Replying to "what about https://g..." Oh yeah, that would be good to include too
- Nico Flaig
5115 has no consensus yet though?
- spencer
Silence = yes?
- Eitan Seri-Levi
It’s not a huge lift for us but would delay our devnet readiness by a bit
- marc | wolovim
Potuz in #interop channel: "I don't think I will be able to make it to ACDT so I'll leave my vote here to include 8045 in devnet 5”
- Nico Flaig
potuz said yes
- danceratopz
- Justin Traglia
Reacted to "Potuz in #interop ch..." with 🙏
- James He
Potuz voted yes to include 8045
- Justin Traglia
Fwiw I also like 8045 & support including it.
- Dmitrii Shmatko
Reacted to "Fwiw I also like 804..." with 👍
- Nico Flaig
8045 doesn't matter
- danceratopz
---------- Spec Changes: 7928 non-empty change set ---------- 📋 EIP-7928 BAL - SlotChanges must have >=1 StorageChange (no empty sets) <https://github.com/ethereum/EIPs/pull/11750>
- Ben Adams
targetGasLimit in PayloadAttributesV4 isn't compatable with the BAL devnets?
- danceratopz
---------- Spec Changes: 7928 state-access decoupling ---------- 🧩 EIP-7928 - decouple state-access validation from EIP-2929 (stack/balance checks to pre-state; warming+BAL paired with post-state cost) <https://github.com/ethereum/EIPs/compare/master...nerolation:EIPs:toni/state-access> R&D discussion <https://discord.com/channels/595666850260713488/1506543443172921464/1509797201340137544>
- Toni Wahrstätter
BAL / Change charge order: https://github.com/ethereum/EIPs/compare/master...nerolation:EIPs:toni/state-access
- Toni Wahrstätter
That's how it looks in EELS: https://github.com/ethereum/execution-specs/pull/2900
- Dragan Rakita
We need to change it too
- Karim T. (matkt)
I will prefer to test to implement it and see the impact before . But for the moment seems to not be a simple change (besu)
- Dragan Rakita
Yeah
- danceratopz
Reacted to "I will prefer to tes..." with 👍
- danceratopz
---------- Spec Changes: 8037 calldata floor (merged) ---------- 🧾 EIP-8037 - calldata floor accounting alignment & call-frame refill clarification <https://github.com/ethereum/EIPs/pull/11706> ---------- Spec Changes: 8037 auth refunds ---------- 🧾 EIP-8037 - expand EIP-7702 authorization gas refunds (5-rule scheme) TLDR: want it or not? added complexity / attack vectors? <https://github.com/ethereum/EIPs/pull/11715>
- danceratopz
---------- Spec Changes: 2780 ---------- ⛽ EIP-2780 - remove self-transfer & precompile carve-outs TLDR: open Q - charge TRANSFER_LOG_COST for self-transfers (no 7708 log)? one cost or Code-vs-No-Code split? <https://github.com/ethereum/EIPs/pull/11735>
- Dragan Rakita
I am fine with the log part.
- Toni Wahrstätter
What about 7997?
- Toni Wahrstätter
the factory precompile
- my side I like the following below, and adding all EL changes and new EL EIPs
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.
- Stefan Starflinger
7997 is marked https://eips.ethereum.org/EIPS/eip-7773 as CFI
- danceratopz
https://github.com/ethereum/EIPs/blob/master/EIPS/eip-7997.md
- Dragan Rakita
Ship it I guess
- Stefan Starflinger
Reacted to "Ship it I guess" with 👍
- Stefan Starflinger
Replying to "Ship it I guess" 🛳️
- Etan (Nimbus)
I mean, the Q about 7688 is, how do we get prysm/teku/lighthouse/grandine to become ready? we keep pushing it to “next devnet”
- Dragan Rakita
Reacted to "🛳️" with 🛳️
- spencer
Replying to "Ship it I guess" 🛳️
- spencer
Reacted to "🛳️" with 🛳️
- danceratopz
https://github.com/ethsteel/pm/blob/main/complexity_assessments/EIPs/EIP-7997.md
- Ben Adams
Replying to "Ship it I guess" 🛳️
- Ben Adams
Reacted to "Ship it I guess" with 👍
- Louis
Replying to "the factory precompi..." It is a system contract?
- spencer
Reacted to "🛳️" with 🛳️
- spencer
Reacted to "🛳️" with 🛳️
- Stefan Starflinger
Reacted to "🛳️" with 🛳️
- spencer
Reacted to "Ship it I guess" with 🛳️
- Stefan Starflinger
Reacted to "🛳️" with 🛳️
- Stefan Starflinger
Reacted to "🛳️" with 🛳️
- Dragan Rakita
Replying to "the factory precompi..." It is bytecode that is deployed at 0x12 address
- Toni Wahrstätter
I'd keep 7997 in CFI for now and ask clients to double-check if we actually need it. E.g. this one: https://safefoundation.org/blog/safe-research-meet-erc-7955-no-private-key-required We shouldn't include it just because it's small but only if it makes much sense.
- danceratopz
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.
- Dragan Rakita
For repricing, I would be fine to change it to anything, just to test the change and later update it when we get better number. Just so we are not blocked by this.
- spencer
optimizations + benchmarking on bal-devnet-7 all new el eips and eip changes on glam-denvet-6 ELs should prio optimizations over glam-devnet-6 EIPs/changes?
- Maria Silva
Also would like to hear from erigon, geth and besu as they are still working on optimizations
- Toni Wahrstätter
benchmarking/optimizations > more eips
- Stefan Starflinger
Reacted to "benchmarking/optimiz..." with 👍
- Mario Vega
Reacted to "benchmarking/optimiz..." with 👍
- Stefan Starflinger
sounds great
- Dragan Rakita
Reacted to "Also would like to h..." with ➕
- Stefan Starflinger
7997?
- felix (eest)
Reacted to "benchmarking/optim..." with 👍
- Mario Vega
I’d prioritize SELFDESTRUCT burn removal, because it simplifies other things
- Dragan Rakita
Reacted to "I’d prioritize SELFD..." with ➕
- Toni Wahrstätter
Reacted to "I’d prioritize SELFD..." with ➕
- Dragan Rakita
Lets focus all EL changes to glam-devnet-6
- spencer
Reacted to "Lets focus all EL ..." with ➕
- Ben Adams
Reacted to "Lets focus all EL ch..." with ➕
- Mario Vega
I meeant over 7997, sorry for confusion
- Edgar | ethrex
Reacted to "Lets focus all EL ..." with ➕
- Stefan Starflinger
Reacted to "Lets focus all EL ch..." with ➕
- Stefan Starflinger
okay then lets postpone it
- spencer
Replying to "Lets focus all EL ..." For clients that have optimizations complete, can you work on this so we have someone to test against?
- Ben Adams
More important to have a working glamsterdam devnet than adding more EL changes to it ahead of the other EL changes?
- Toni Wahrstätter
optimizations > selfdesctruct burn > 7997
- Mario Vega
Reacted to "optimizations > self..." with 👍🏼
- Andrew Ashikhmin
Reacted to "optimizations > self..." with 👍🏼
- Stefan Starflinger
Reacted to "optimizations > self..." with 👍🏼
- Dragan Rakita
Lets list all needed changes for glam-devnet-6 so people know what is needed
- Dragan Rakita
Reacted to "optimizations > self..." with 👍🏼
- Dragan Rakita
Replying to "Lets focus all EL ch..." Yeah already have some version of all of this
- Dragan Rakita
Replying to "Lets focus all EL ch..." Intrinsic gas is maybe only thing missing
- spencer
Reacted to "Yeah already have ..." with 🔥
Call summary
Targets
- •glamsterdam-devnet-5 launch - as early as tomorrow (June 2), pending alpha.9 spec release - 00:13:38
- •bal-devnet-8 / glamsterdam-devnet-6 decision deferred to ACDT next week - 00:58:35
Decisions
- •EIP-8045 included in glamsterdam-devnet-5; all CL teams in agreement - 00:19:54
- •glamsterdam-devnet-4 abandoned; not recovered - 00:20:11
- •glamsterdam-devnet-5 is CL-only (EIP-8045 + stability fixes); all EL changes target glamsterdam-devnet-6 - 01:02:06
- •Priority order for EL work: benchmarking/optimizations > self-destruct burn (EIP-8246) > EIP-7997 - 01:00:51
- •Gas repricing breakout calls ended; all discussion moves to ACDT - 00:06:29
Highlights
- Spec Changes:
- ·EIP-7928 spec change merged: SlotChanges must contain at least one StorageChange (no empty sets) - 00:21:15
- ·EIP-7928 state-access decoupling PR open: charge 100 gas pre-depth/balance checks, then remaining 2500 after; Reth proposed, most clients need changes - 00:23:52
- ·EIP-8037 auth refund expansion PR open (5-rule scheme): refund state gas for invalid or net-zero authorization list entries - 00:31:14
- ·EIP-2780 simplification pending: remove self-transfer and precompile carve-outs; awaiting benchmark results before merge - 00:38:29
- Testing Progress:
- ·bal-devnet-7 stable with 100% EVM fuzzing participation; clients iterating on benchmarking - 00:07:33
- ·EIP-7975 (eth/70) and EIP-8159 (eth/71) confirmed implemented across all EL clients - 00:08:43
- ·Final test release tests-bal@v7.3.0 will be cut once EIP-8037 lands in forks/amsterdam - 00:11:14
- Fork Status And Schedule:
- ·glamsterdam-devnet-4 abandoned following ePBS split at ~slot 22800; root cause: should_build_on_full / PTC timing bug - 00:12:51
- ·EIP-8045 (exclude slashed validators from proposing) included in glamsterdam-devnet-5; all CL teams confirmed ready or near-ready - 00:19:54
- ·Consensus: glamsterdam-devnet-5 CL-only; all EL changes (EIP-8246, EIP-2780, EIP-8038, EIP-7954, EIP-7997) deferred to glamsterdam-devnet-6 - 01:02:06
- ·Benchmarking optimizations on bal-devnet-7 are the primary EL blocker; must complete before final gas cost numbers can land - 00:55:17
Action Items
- •All EL teams - EL teams review EIP-7928 state-access decoupling PR before ACDE Thursday - 00:30:19
- •danceratopz - danceratopz to open ETH R&D thread outlining glamsterdam-devnet-5 scope for async consensus - 01:05:30
Key decisions
EIP-8045 included in glamsterdam-devnet-5; all CL teams in agreement
glamsterdam-devnet-5glamsterdam-devnet-4 abandoned; not recovered
DFIglamsterdam-devnet-5 is CL-only (EIP-8045 + stability fixes); all EL changes target glamsterdam-devnet-6
glamsterdam-devnet-5CL-only, EL changes in glamsterdam-devnet-6Gas repricing breakout calls ended; all discussion moves to ACDT
Gas repricing breakout calls ended