upgrade: upgrade xlayer-reth with reth v2.1.0 and op-reth v2.1.0#270
upgrade: upgrade xlayer-reth with reth v2.1.0 and op-reth v2.1.0#270
Conversation
Bump reth to okx/reth@056289cd (v2.0.0), align all alloy/revm/op-alloy deps to versions used by reth, bump optimism submodule, drop obsolete op-alloy crates.io patches, and bump toolchain to 1.94.0. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
sieniven
changed the title
Migrate xlayer-reth call sites for the upstream reth v2.0.0 and op-reth v2.1.0 upgrade: - Replace removed `PayloadBuilderAttributes` trait usage with `OpPayloadBuilderAttributes` struct + `OpPayloadAttrs` engine API wrapper; flatten nested `payload_attributes.X` field accesses. - Drop the removed `TaskSpawner` / `TokioTaskExecutor` generics in flashblocks builder, generator, handler, RPC subscription, and monitor in favor of concrete `reth_tasks::TaskExecutor`. - Rewrite `DefaultPayloadBuilder` `PayloadBuilder` impl to mirror upstream `OpPayloadBuilder` (`Attributes = OpPayloadAttrs`, convert to internal builder attrs in `try_build`). - Plumb `runtime: reth_tasks::Runtime` through bin/tools commands and the new `import_blocks_from_file` arg. - Add `cache_for`, `sparse_trie_handle_for`, and a separate `WaitForCaches` impl on `XLayerEngineValidator` to satisfy the new `EngineValidator` trait surface; return the new `(ExecutedBlock, Option<Box<ExecutionTimingStats>>)` tuple. - Update `OpFlashblockPayloadMetadata` construction for the new `Option<BTreeMap>` fields plus `access_list`. - Migrate validator state-root pipeline to the new payload_processor surface: keep `Arc<TrieUpdates>` end-to-end into `DeferredTrieData::pending`, switch from the now-private `ChangesetCache::insert` to `register_pending`/`resolve`, set `ExecutionEnv.gas_used`, and box the `OnStateHook` from the optional state hook. - Preserve `WithEncoded` bytes through `flashblock_tx_iterator` so the prewarm path uses `FromTxWithEncoded` instead of recomputing `tx.encoded_2718()` per transaction via `FromRecoveredTx`. - Add `PostExec` arms for `OpReceipt` and `OpTypedTransaction`, pairing them with the existing `Deposit` arm in `Signer::sign_tx` (system transactions, never user-signed). - Implement `PoolTransaction::consensus_ref` on `MockFbTransaction` mirroring upstream's `unimplemented!()` mock pattern. - Replace the removed `set_state_clear_flag` call (Spurious Dragon empty-account handling is always on in revm 12.0). - Derive next-block number from `provider.last_block_number() + 1` in `monitor::handle` since `Events::Attributes` no longer carries parent hash. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
@claude review my diff for me, with the new reth v2.0.0 version upgrade |
XLayer-Bot
left a comment
XLayer-Bot
left a comment
There was a problem hiding this comment.
Overall this is a clean and well-executed adaptation to the reth v2.0.0 / op-reth v2.1.0 API surface. The mechanical changes — flattening EthPayloadBuilderAttributes into OpPayloadBuilderAttributes, removing the OpAttributes bound, concretising Tasks type params to TaskExecutor, propagating the (ExecutedBlock, timing) tuple from validate_payload/validate_block, wrapping TrieUpdates in Arc, updating OpFlashblockPayloadMetadata fields to Option<…>, adding OpReceipt::PostExec coverage, and replacing [patch.crates-io] redirects with native path deps for op-alloy — are all correct and faithful to upstream.
Three minor items inline below.
The Rust badge was bumped in the v2.0.0 upgrade commit but the Reth badge link/version was missed. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
@claude review again |
XLayer-Bot
left a comment
XLayer-Bot
left a comment
There was a problem hiding this comment.
All three previously raised concerns have been addressed:
README badge — updated to reth-v2.0.0 in 33be64b. ✓
monitor/handle.rs last_block_number() — author confirmed this is an acceptable approximation for monitoring metrics, and the change was unavoidable given that reth v2 no longer exposes the parent hash in the Events::Attributes variant. ✓
builder_tx.rs enveloped_tx: Some(Bytes::new()) — author verified that the previous OpTransactionRequest::try_into_tx_env impl (alloy-evm 0.26.3) already produced enveloped_tx: Some(Bytes::new()), confirmed by an upstream unit test. The new explicit construction is byte-identical to the old behavior. ✓
The rest of the reth v2 / op-reth v2.1.0 API migration looks mechanically correct: OpTxEnvConverter adoption, TaskExecutor concretisation, spawn_blocking_named additions, OpReceipt::PostExec arm, Arc<TrieUpdates> wrapping, deferred changeset provider creation, and the set_state_clear_flag removal (safe — reth v2's State::builder().with_bundle_prestate(bundle) inherits the correct Spurious Dragon flag from the accumulated bundle, making the explicit call redundant).
Bump okx/reth pin to ccbef3f and update deps/optimism submodule. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
🔧 CI running for |
|
❌ XLayer Reth CI failed for commit
|
|
🔧 CI running for |
|
❌ XLayer Reth CI failed for commit
|
|
🔧 CI running for |
|
❌ XLayer Reth CI failed for commit
|
|
🔧 CI running for |
|
✅ XLayer Reth CI passed for commit
|
sieniven
force-pushed
the
niven/reth-v2.0.0
branch
from
12c782c to
c9e40fd
Compare
- Add [patch."https://github.com/paradigmxyz/reth"] in workspace root so reth-* crates pulled transitively via deps/optimism/rust path crates resolve to the same okx/reth rev pinned in [workspace.dependencies], collapsing the lockfile to a single reth source. - Bump deps/optimism submodule to 73460092 (sync upstream). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…VE patch The local fork was at 1.6.3 from when the workspace shipped reth v1.11.3. The reth v2.0.0 alignment bumped alloy-rpc-types-engine to 1.8.x, leaving the 1.6.3 fork outside semver range and silently unused — jsonwebtoken fell back to the vulnerable 9.3.1 from crates.io. Replace the fork's contents with upstream alloy-rpc-types-engine 1.8.3, re-apply the jsonwebtoken pin to 10.3.0 with the rust_crypto backend, and add the missing [patch.crates-io] header above the security patches so both the alloy-rpc-types-engine and libp2p-yamux overrides actually take effect. Cargo.lock now resolves jsonwebtoken to 10.3.0, libp2p-yamux to the local fork (no yamux 0.12.x), and emits no patch-unused warnings. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Bump okx/reth rev from ccbef3f8 to 2de4f640 across both [workspace.dependencies] and [patch."https://github.com/paradigmxyz/reth"] so the workspace stays on a single reth source. Cargo.lock fully migrated to the new rev (103 refs, no stragglers on the old one). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
@claude review again for me |
XLayer-Bot
left a comment
XLayer-Bot
left a comment
There was a problem hiding this comment.
All three previous review threads are resolved — the badge fix, the monitoring approximation (accepted trade-off), and the enveloped_tx behavior (verified byte-identical to v1.11.3). The four new commits are clean:
chore: fmt— pure formatting, no logic change.chore(deps): patch reth to okx fork— correctly adds[patch."https://github.com/paradigmxyz/reth"]so transitive reth deps pulled in bydeps/optimismresolve to the sameokx/rethrev, collapsing the lockfile to a single source. The op-alloy / alloy-op-* crates were removed from[patch.crates-io]because they are now declared as direct workspace path deps, which takes precedence without a patch entry — verified in the lock file.fix(deps): rebase alloy-rpc-types-engine fork onto 1.8.3— restores the CVE-2026-25537 jsonwebtoken security patch that was silently unused after the 1.6.3→1.8.x bump (the[patch.crates-io]entry didn't match the new semver range). The manualDeserializeimpls for V2/V3/V4/InputV2/EnvelopeV4 are correct: the derivedSerializeretains#[serde(flatten)]on the inner payload fields, and theHelperstructs in each customDeserializeread the same flat camelCase fields. Round-trip tests validate this. TheJwtSecretDebug name typo fix from upstream 1.7.4 is also included.chore(deps): switch okx/reth tip— bumps the rev to pick up the storage v2 static-file fix, keeping both[workspace.dependencies]and the new[patch."https://github.com/paradigmxyz/reth"]on the same rev. Verified the lock file has a single reth source.
Replace Runtime::test() (test helper with 2 worker threads, 2 rayon threads, 16 max blocking tasks) with a production-grade Runtime built via RuntimeBuilder using RuntimeConfig::default() and attached to the existing #[tokio::main] handle. This gives full CPU-derived rayon pool sizing for import/export/gen-genesis/legacy-migrate operations without spawning a second tokio runtime. Addresses PR #270 review comment on bin/tools/main.rs:63. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
sieniven
changed the title
Align workspace deps to reth v2.1.0 (okx/reth@044b173): - reth-primitives-traits 0.1.0 → 0.3.0 - revm 36.0.0 → 38.0.0; op-revm 17.0.0 → 19.0.0 - alloy-evm 0.30.0 → 0.33.0; alloy-op-evm 0.30.0 → 0.31.0 - alloy-hardforks 0.4.5 → 0.4.7 - alloy-{consensus,contract,eips,genesis,json-rpc,network,provider, rpc-client,rpc-types-{beacon,engine,eth},serde,signer-local, transport,transport-http} 1.8.2 → 2.0.0 - tokio 1.44.2 → 1.51.1; tokio-tungstenite 0.26.2 → 0.28.0 - metrics 0.24.1 → 0.24.0 - README badge: reth v2.0.0 → v2.1.0 Drop the local alloy-rpc-types-engine fork: the v1.8.3 jsonwebtoken CVE patch is no longer needed since alloy-rpc-types-engine 2.0.0 ships with jsonwebtoken 10.3.0 upstream. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- alloy-consensus 2.0 added block_access_list_hash and slot_number to Header; populate as None in flashblocks builder and assemble paths (matches op-reth/crates/evm/src/build.rs) - alloy-rpc-types-eth 2.0 added block_timestamp to TransactionInfo; use ..Default::default() to match reth v2.1.0 RPC convention - alloy-network 2.0 split TransactionRequest::build off into the NetworkTransactionBuilder trait; import it in e2e test utils - revm 38 deprecated ExecutionResult::gas_used / Gas::used after the EIP-8037 state-gas split; switch flashblocks builder, context, and handler to tx_gas_used() 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
5 participants
Summary
alloy-rpc-types-enginefork: the v1.8.3 jsonwebtoken CVE-2026-25537 patch is no longer needed becausealloy-rpc-types-engine2.0.0 ships withjsonwebtoken10.3.0 upstream🤖 Generated with Claude Code