Repository navigation
op-reth: fix reth v2.6.0 review findings - #23118
Merged
joshklop merged 5 commits intoOct 12, 2026
Merged
Conversation
sebastianst
force-pushed
the
seb/reth-v2.6.0-review-fixes
branch
from
October 1, 2026 08:57
1e1fd39 to
7fb53f9
Compare
joshklop
force-pushed
the
joshklop/update-reth-v2.6.0
branch
from
October 10, 2026 22:11
e8453a0 to
12d73f7
Compare
joshklop
force-pushed
the
seb/reth-v2.6.0-review-fixes
branch
from
October 11, 2026 22:03
a1cbb76 to
9dedb67
Compare
With state masking, reth persists blocks through the Finish checkpoint while its hashed-state and trie tables lag at the partial state trie frontier until `op-reth node` repairs them on launch. `proofs init` copied those tables as the tip's state, seeding a proofs store that matches no block after an unclean shutdown or a live-node snapshot. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Stop serving `debug_getRawBlockAccessList`, which reconstructs block access lists through the same untested OP execution path as the eth methods, and refuse to launch with `--rpc-cache.prewarm-bals` or `--rpc-cache.cache-computed-bals`: prewarming never stops on OP chains, and cached lists would position tracing state. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
joshklop
force-pushed
the
seb/reth-v2.6.0-review-fixes
branch
2 times, most recently
from
October 11, 2026 22:58
a9acc1b to
41a5488
Compare
…ckports Pin reth to a fork commit that adds upstream fixes on top of the CanonicalInMemoryStateProvider pin: - reth#27129: the `OverlayStateProvider` storage root and storage proofs read trie nodes through the overlay, and historical proofs complete the overlay to the Finish checkpoint. With state masking, the database storage trie omits nodes overwritten by the masking suffix, so the Isthmus withdrawals root (the L2ToL1MessagePasser storage root) was computed from stale cached child hashes, both when building blocks and when validating them. A regression test covers masked and unmasked tries over both storage versions. - reth#27634, the follow-up to #27129: use trie changesets for every overlay. Without it, #27129 makes historical proofs and witnesses for blocks below the in-memory window rehash from the leaves, which takes minutes and hits the read-transaction timeout. - reth#27247 and reth#27429: validate safe and finalized before applying a forkchoice reorg, and stop reporting VALID for a reorged-out persisted head. - reth#27252: the state root task persisted accounts emptied under EIP-161 as all-zero `HashedAccounts` rows instead of deleting them. The block's root is still right, but a later full recomputation of that trie branch diverges and halts the node. - reth#27832: the walker used by the startup Merkle unwind, which repairs a masked state trie after an unclean shutdown, could leave stored branch nodes that their parent no longer references. The regular walker skips them, so the repair's root check passes, but proof_v2 reads them and the engine can then compute a wrong state root. The fork branch follows the maintenance procedure: the upstream release, the retained patches oldest first, then the CI commit. The backports change no symbol a mirror covers, so every reth mirror tag advances to the new pin unchanged. Co-authored-by: Sebastian Stammler <seb@oplabs.co> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`proofs init` reads the chain tip it anchors to before it opens the copy transaction, and it opens the database read-only, so a node still running on the datadir can persist past that tip in between. With state masking off, the lagging-trie check then passes and the proofs store anchors the earlier block over tables from a later one. Refuse when the copy transaction's Finish checkpoint no longer matches the anchor, and tell operators to stop any node on the datadir. Also drop a dev-dependency on reth-chain-state, which is now a regular dependency. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
op-reth's CLI already rejects every block access list option, so the add-ons check only guards code that builds a `NodeConfig` without that CLI. Say so, point new upstream flags at `DENIED_ARGS`, and track the auth-port `EngineEthApi`, which also declares the unserved methods. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
joshklop
force-pushed
the
seb/reth-v2.6.0-review-fixes
branch
from
October 12, 2026 00:33
41a5488 to
16aa7fc
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This fixes the blocking findings from my review of #23046. Each one is a regression that the reth v2.6.0 bump introduces. Stacked on #23046.
withdrawalsRootinto a block header, and a verifier could reject a valid block withL2WithdrawalsRootMismatch. The cause is upstream:OverlayStateProvider::storage_rootignored the overlay trie nodes. This PR pins reth to perf(trie): backport reth#27129 overlay storage roots and historical proofs reth#11, which backports the upstream fix (reth#27129, released in v2.7.0); that backport also fixes historical proofs over masked rows. A regression test, masked and unmasked over both storage versions, fails on rust!: update reth to v2.6.0 #23046.op-reth proofs init: init now refuses a datadir whose state trie lags its tip, or where reth's repair of it was interrupted, which happens after an unclean shutdown or on a snapshot of a running node. Before, init seeded a proofs store that matched no block. With backfill, that then failed withStateRootMismatch; without backfill, the node silently served wrong historical proofs. The error tells operators to start the node once and stop it cleanly.debug_getRawBlockAccessListis no longer served. op-reth now refuses to start with--rpc-cache.prewarm-balsor--rpc-cache.cache-computed-bals: on OP, prewarming re-executes every block forever, and it feeds the untested block access list reconstruction into tracing.flashblock_receipts_streamreturns logs for filters with an address or topic again. Since the bump, it returned none for them.Before merge: keep this as a draft until ethereum-optimism/reth#11 and #13 merge, then re-pin to the merged commit. The current pin is reachable only from the fork PR's branch.
Not addressed:
RethTrieStorageLayoutconversion, named test tuples, and a network-args helper.transaction_index/log_index.🤖 Co-created with Claude Opus 5.5