Repository navigation
Description
Test, test-support and DSL code still calls op-geth-only symbols for which op-core/* already has replacements. The plan (docs/ai/opgeth-decoupling.md §1, §13, §15) says these sites "ride §13", but none of the #20265 sub-issues covers them: those move L2 RPC access onto apis.EthClient, not symbol uses.
Some of this debt was added after the op-core helpers existed (derive tests, the sdm acceptance tests, kona proof tests). The opgeth-ratchet CI check (#23407) stops new uses. Its allowlist, ops/scripts/opgeth-ratchet/allowlist.txt, lists the current sites with reasons starting #23405, and just opgeth-ratchet-update drops entries as they are fixed.
Scope, each a swap onto op-core/{types,fees,eip1559,params} or optypes.Receipt:
- op-devstack DSL —
dsl/fjord_fees.goanddsl/operator_fee.gocompute expected fees with op-geth's cost functions and read geth receipt fee fields. - op-acceptance-tests — jovian, fjord, base, batcher/throttling, sdm and the
sdmtestharness that external suites import (IsDepositTx,RollupCostData,EncodeJovianExtraData,DepositTxType,PostExecTx, receipt fee fields), fork-time reads insync_tester_hfsandinterop/upgrade, and the interop load test'score.ErrTxFilteredOutmatch (follows the txinclude change in opgeth-decoupling: swap or delete remaining non-test fork-symbol uses #23406). - op-node —
rollup/derivetests androllup/derive/test(DepositTx,PostExecTx,Mint,SourceHash,IsSystemTx,EncodeOptimismExtraData,NewBlock/DefaultBlockConfig). - op-service —
testutils, and thebgpo,txplan,sources,txincludetests (NewBlock/DefaultBlockConfig,DepositTx, receipt fields, Fjord cost functions). - op-batcher —
batchertests (NewBlock/DefaultBlockConfig). - op-interop-filter, op-interop-mon —
filterandmonitortests readparams.InteropCrossL2InboxAddress→predeploys.CrossL2InboxAddr. - op-e2e —
system/{bridge,fees}(DepositTx,NewIsthmusSigner, eip1559 and cost functions),actions/{derivation,upgrades}cost-function uses, and the in-process L1 miner and replica inactions/helpers(l1_miner.go,l1_replica.go), which stay in-process after opgeth-decoupling: migrate op-e2e/actions onto op-reth-test-engine (subprocess EL) #21196 and use fork signatures (NewEVMBlockContext,eip1559.CalcBaseFee,NewBlock) and options (PrecompileOverrides,RollupDisableTxPoolGossip). - cannon —
mipsevm/testutil/evm.goruns the MIPS contracts on a plain L1 (mainnet-config) EVM but calls the fork'score.NewEVMBlockContext(header, chain, author, config, statedb); upstream takes(header, chain, author). The extra arguments only feed the OP L1-cost and operator-fee functions, so the swap just drops them (same for the op-e2e L1 miner). - contracts-bedrock go-ffi —
scripts/go-ffi/utils.goencodes deposits for the Solidity tests withtypes.DepositTx. - kona proof tests —
rust/kona/tests/proofsGo tests (DepositTxType,IsDepositTx,IsSystemTx,NewIsthmusSigner, eip1559 and L1-cost helpers). opgeth-decoupling: migrate the kona proof harness off the in-process op-geth L2 EL #21905 moves the harness off the in-process EL, but not these symbol uses. kona/tests/proofs: read the L2 chain over RPC #23417 swaps the uses on the L2-read paths it rewrites (optypes.DepositTxType,optypes.UnmarshalDepositTx,fees.TxRollupCostData,fees.L1CostFjord). Re-derive the rest after it merges. - op-chain-ops/pkg/sdm tests (
DepositTxType,PostExecTx).
Out of scope: op-e2e/opgeth (deleted by #20275), actions/helpers/engineapi (#21196), the in-process op-geth L2 EL (#21744), and the op-core differential tests, which go at the cutover (#20266).
Done when the ratchet allowlist has no entries for these packages.
🤖 Co-created with Claude Opus 5.5
Activity
sebastianst commentedon Oct 26, 2026
🤖 Opus 5.5 says: One trap in the
eip1559.CalcBaseFeeswap: kona'soperator_fee_test.go#L214calls the fork'sCalcBaseFee(config, parent, time)on the L2 config.op-core/eip1559has no OPCalcBaseFee(only the extraData encode/decode/validate helpers), and upstream'sCalcBaseFee(config, parent)compiles but computes the Ethereum base fee: denominator 8, elasticity 2, no Holocene extraData params, no Jovian minimum base fee. So the swap shouldn't take upstream's signature. Either read the next base fee from op-reth, or add an OPCalcBaseFeetoop-core/eip1559with a differential test.sebastianst commentedon Nov 12, 2026
🤖 Fable says: Reopening: #24090 moved the go-ffi helper to op-core types, but the remaining test and DSL swaps land with #23864 and #23863. #23864 closes this.