Repository navigation
Glamsterdam (Sepolia Amsterdam) L1 compatibility: op-geth pin, L1 header parser, batcher gas floor - #52
Merged
Conversation
…pt to its API
Point the go-ethereum replace at hemilabs/op-geth f5c491e251ce (branch
glamsterdam-l1-header-compat, on top of hemi 6ceee9fc3), which decodes and
hash-verifies post-Glamsterdam Sepolia L1 headers. This pin is load-bearing on
its own: op-node's L1 head signal (eth_getBlockByNumber latest / newHeads ->
*types.Header -> Hash()) and op-batcher's L1 tip hash headers through op-geth's
types.Header, which the --l1.trustrpc flag does not bypass, so only an image
built from this pin tracks Sepolia past the fork.
go.mod carries the two indirect requirements the new op-geth needs. go.sum
carries their entries plus the two entries for the op-geth pseudo-version,
computed from the local checkout with the same module-zip code go uses (the
method was validated by reproducing the previous pin's real hashes). The
op-geth commit is not on GitHub yet: push that branch unchanged (amending or
rebasing it changes the pseudo-version and the hashes), then run `go mod tidy`
and confirm it produces no diff; a checksum mismatch would mean the pushed
commit differs from f5c491e251ce.
The previous pin (e92aa4e51692, January 2026) predates two op-geth API changes.
Every call site, including tests, is adapted rather than vendoring old copies:
- consensus/misc/eip1559.ValidateOptimismExtraData / ValidateHoloceneExtraData /
ValidateJovianExtraData take the block gas limit (op-geth ethereum-optimism#89 rejects an
elasticity larger than the gas limit, and a zero denominator/elasticity pair).
op-node must apply the same rule when validating attributes against blocks,
otherwise it could accept extraData that op-geth rejects, so the gas limit of
the object whose extraData is validated is passed through
checkExtraDataParamsMatch and the payload-to-SystemConfig path. Test fixtures
that relied on arbitrary elasticity bytes or on a zero-elasticity attribute
set are updated to realistic values and to the stricter op-geth messages (the
blocks are rejected either way); cases for the translated-zero mismatch and
for the gas-limit rule are added.
- state.StateDB.SetCode/SetBalance take a tracing reason; core.ChainContext
embeds consensus.ChainHeaderReader; ethdb.Database gained AncientBytes;
core.NewBlockChain's Hemi signature. Mechanical updates in op-chain-ops,
cannon, op-wheel, op-sync-tester and op-program.
Against the new pin `go build ./...` fails in a strict subset of the packages
that already fail on hemi with the old pin (check-jovian, op-simulate,
sync-tester, fakebeacon and op-wheel/cheat now compile); the shipped binaries
(op-node, op-batcher, op-proposer) build, and no package that built before
stops building.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ation
Sepolia activates Glamsterdam (EL fork Amsterdam) at timestamp 1791294816
(2026-10-06 13:53:36 UTC). Together with the op-geth pin bump, this fixes the
two remaining things in this repository that break at that block:
- op-node rebuilds every L1 header it fetches from the RPC's JSON fields and
rejects it if the recomputed hash differs from the one the node reported.
Post-fork headers carry blockAccessListHash (EIP-7928) and slotNumber
(EIP-7843) after requestsHash; without them the recomputed hash is wrong and
L1 head tracking and derivation stop at the first post-fork block. Map both
fields in RPCHeader and CreateGethHeader (port of
ethereum-optimism#21920, shipped upstream in op-node v1.19.5). The
tests pin a header produced by go-ethereum v1.17.7: the hash verifies with
both fields, fails without either, and the same header without the fields
hashes as a 21-field pre-fork header. Verified separately against all 3,277
official execution-spec-tests Amsterdam blockchain fixtures (53,300 headers,
zero mismatches).
- op-batcher sets calldata batch transactions' gas limit from
core.FloorDataGas, which op-geth deliberately keeps at the pre-Amsterdam
(L2) rule. Sepolia will reject those transactions under EIP-2780/EIP-7976
(15,000 base plus 64 gas per calldata byte, zero bytes included). Use the
larger of the pre- and post-Amsterdam floors (port of
ethereum-optimism#21921, shipped upstream in op-batcher v1.17.0),
which is valid on both sides of the fork. Hemi's batchers currently run
with --data-availability-type=blobs, whose transactions carry no calldata
(gas limit 21,000, above Amsterdam's 15,000 intrinsic cost), so this matters
only if the DA type is ever switched; the auto-DA cost model is left on
pre-Amsterdam pricing for the same reason.
Receipts, logs, transactions, blob retrieval and the L1 blob base fee are
unaffected by the fork and need no change. Rebuild the image and roll it out to
op-node, op-batcher and op-proposer before the activation; --l1.trustrpc is not
a substitute, because the L1 head signal path hashes headers regardless.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…s' tests compile NewAttributesHandler, NewL1Retrieval, DataSourceFactory.OpenData and NewRPCReceiptsFetcher gained a hemitrap flag without their tests being updated, so the op-node/rollup/attributes, op-node/rollup/derive and op-service/sources test packages did not compile. Pass the flag (disabled, preserving the original test behaviour) so the Glamsterdam tests in those packages can run. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
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
1 participant
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.
Why
Sepolia activates Glamsterdam (EL fork Amsterdam) at timestamp 1791294816 (2026-10-06 13:53:36 UTC). Post-fork L1 headers carry
blockAccessListHashandslotNumberafterrequestsHash. op-node recomputes every L1 header hash from RPC JSON and from op-geth'stypes.Header(L1 head signal, batcher L1 tip) and rejects mismatches, so L1 tracking and derivation stop at the first post-fork block on the current image.--l1.trustrpcis not a workaround: the head-signal path hashes headers regardless.What (three commits)
glamsterdam-l1-header-compat, tagglamsterdam-l1-header-compat-20261005), which decodes and hash-verifies Amsterdam headers, and adapt every call site to the op-geth API changes since the January pin. In particular op-node now passes the block gas limit into extraData validation, matching the consensus rule Hemi's op-geth already enforces (op-geth Error Handling and Logging around OVM -> EVM Tx hash map ethereum-optimism/optimism#89), so op-node and op-geth cannot disagree on a block.RPCHeader/CreateGethHeader(port of acceptance: support newer L1 forks ethereum-optimism/optimism#21920, op-node v1.19.5) and use a calldata gas floor valid on both sides of the fork (port of acceptance: test Glamsterdam L1 transition ethereum-optimism/optimism#21921, op-batcher v1.17.0). Hemi's batchers run with blobs, so the floor only matters if DA is ever switched.Relation to #50: the API adaptations are re-done by adapting call sites instead of vendoring; the #50 pin (a mid-ethereum-optimism#110 commit without the access-list hash), the txmgr floor formula (under-counts zero bytes), the pre-flight removal, the L1 tracker cache removal and the
BatchPastcase are deliberately not taken (reasons in the review notes).Verification
go mod tidyis a no-op andgo mod verifypasses against the pushed op-geth commit; the shipped binaries build in plain module mode.hemi(three previously uncompilable test packages now compile and pass).Rollout
Merge, let CI build the image, and roll it out to op-node, op-batcher and op-proposer before 2026-10-06 13:53:36 UTC. No flag changes. The paired L2 EL must be
hemiat 6ceee9fc3 or later.🤖 Generated with Claude Code