Repository navigation
deps: update reth from main (2026-11-17) - #9371
Conversation
|
derek bench preset=tip20 slack |
|
cc @decofe ✅ Benchmark complete: Improvement View job ✅ Bench Comparison: ImprovementRefs: 35d8b93 vs reth-auto-bump Configuration
Tempo Metrics
Builder
Builder details
Validator
Persistence
ObservabilityWarn/Error Logs
Counts by message
|
👁️ Cyclops Security Review✅ Reviewed e769a23 (latest) · Coverage: partial · no open findings · report |
There was a problem hiding this comment.
I reviewed this reth bump and found no bugs. Cargo.toml and Cargo.lock have CODEOWNERS, and the new upstream range changes engine/consensus and trie behaviour, so a human look is still worthwhile.
What was reviewed:
- Every reth git dependency in Cargo.toml, jsonrpsee included, moves from 31d05e1 to 36600d5. Features and default-features flags are unchanged.
- All 110 reth
sourcelines in Cargo.lock point at the same new rev, and none still use the old one. Nothing else in the lockfile changed: no versions, checksums or new packages. - Not done: no build,
zepter, or test run was part of this review, so whether Tempo compiles against the new rev is left to CI. The second commit ("fix: resolve breaking changes") changes no Rust source.
Extended reasoning...
The change only updates the pinned paradigmxyz/reth git rev in Cargo.toml and Cargo.lock. No Rust source changes, and the diff touches no auth, injection or data-exposure code. The diff itself is mechanical and consistent: one rev everywhere, and no version or checksum churn. I deferred because .github/CODEOWNERS assigns both changed files to named owners. Also, the upstream range includes consensus-adjacent changes (BAL hash validation, witness key inclusion), and an owner should check those against Tempo's rule that an active fork's behavior must not change.
c9a7d3b to
790bdeb
Compare
790bdeb to
c2c153b
Compare
c2c153b to
8de9b79
Compare
|
|
cyclops audit note="Review every upstream reth commit included in the revision diff." |
|
cc @decofe Cyclops audit started. Config: env: |
8de9b79 to
2b907dc
Compare
There was a problem hiding this comment.
I reviewed the latest push and found no bugs. It does change consensus gas settlement code owned by CODEOWNERS, so a human look would still be worthwhile, mainly to confirm it matches the upstream evm2 code.
What was reviewed:
intrinsic_gasin crates/evm/src/handler.rs: the per-auth state gas is nownew_account_state_gas() + TxEip7702PerAuthState. Tests still pin 225k, plus 225k for nonce==0 on T4.execute_aanow subtractsstate_refundfromstate_gas_spentaftersettle_transactionreturns, instead of inside evm2. I couldn't read the evm2 source at either rev from this checkout, so I couldn't confirm the old code did the same subtraction at the same point.- The handler tests call
settle_transactiondirectly, so none of them runs the new subtraction with a non-zerostate_refund.
Extended reasoning...
This is an automated update of the reth and evm2 git revs, plus two edits to crates/evm/src/handler.rs to follow evm2 API changes in EIP-7702 intrinsic state gas and AA transaction gas settlement. These paths are consensus-critical gas accounting, but they are not auth, crypto or data-exposure code. Whether total state gas is unchanged on active forks depends on the old and new evm2 internals, which were not available to read here. No test covers the moved refund subtraction with a non-zero refund. CODEOWNERS covers crates/evm/ and crates/hardfork/, so this was not approved automatically.
Automated nightly update of reth dependencies from
paradigmxyz/rethmain branch.Upstream reth changes
31d05e1...f551b68🔗 Amp thread: https://ampcode.com/threads/T-01a1e993-b347-76cd-8ff2-f35efe2c042f
RPC
Engine / Consensus
in_memorymodule so in-memory blocks sit behind a single lock (#28507, #28506, #28543)Trie
Storage / DB
EVM
Txpool
--txpool.enforce-tracked-nonceflag (#28526)Build
Testing
Migrations
🔗 Amp thread: https://ampcode.com/threads/T-01a1e994-010c-74bf-ba4d-8f9ddde55ac8
jsonrpseeentry pinned to the reth repo) from31d05e14to0e57f259,reth-corefrom275ce4cfto5dabd396, and evm2 and its DaniPopes patch entries from66687c18toe681662b.GasParams::eip7702_auth_state_gas()helper. Intrinsic gas now addsnew_account_state_gas()andGasId::TxEip7702PerAuthStateitself.eip7702_auth_state_gas()calls in thegas_params.rstests tonew_account_state_gas(). The expected values (225,000 state, 250,000 total per auth) stay the same.state_refundfield fromGasSettlement, since upstream removed it. It is gone fromprepare_aa,execute_aa, and every test that builds aGasSettlement.execute_aanow subtracts the authorization state-gas refund fromstate_gas_spentitself aftersettle_transactionreturns. Before,settle_transactionreceived it throughstate_refund. The refund goes back to the reservoir, so it should not count as state gas spent.GitHub Workflow