tests: update test chain to amsterdam, regenerate fixtures

[MysticRyuujin]

Bumps the test chain from Osaka+BPO2 to Amsterdam and regenerates all fixtures, following the Osaka bump in #846. This gives the fixtures post-Amsterdam blocks, which the callTracer spec (#855) and the EIP-8037 tracing work (#852) build on.

The hive side is merged: ethereum/hive#1589 added Glamsterdam support to hivechain, and ethereum/hive#1587 added the calltree contract and transaction scenarios. The committed chain regenerates byte-for-byte from current hive master.

Changes

• mkchain.sh: -lastfork amsterdam -length 60. Amsterdam activates at block 57, so the head and two more blocks are post-fork. Also writes the new bal output (per-block access lists, committed under tools/chain/bal as a reference for EIP-7928 work).
• tools/go.mod: go-ethereum pinned to master (86696a8f, 2026-08-28). This picks up the RPC fixes that landed after glamsterdam-devnet-8: eth_estimateGas returns used gas for plain transfers (eth/gasestimator: return used gas for plain transfer estimates go-ethereum#35592) and the debug_traceCall block parameter is optional (eth/tracers: make debug_traceCall block parameter optional go-ethereum#35583). The other post-devnet-8 commits change no fixture output.
• testgen adaptations to Amsterdam:
  • eth_simulateV1 log counts include the EIP-7708 transfer logs.
  • eth_estimateGas: the simple-transfer case now expects 15000. EIP-2780 prices a zero-value transfer to an existing account below 21000, and geth reports the used gas since eth/gasestimator: return used gas for plain transfer estimates go-ethereum#35592.
  • testing_buildBlockV1 and testing_commitBlockV1 (testing: add testing_commitBlockV1 spec and fixtures #801) send slotNumber (EIP-7843) in the payload attributes; the buildBlockV1 schema documents the parameter.
  • the complete-eth-sends simulate case funds its recipient. Post-Amsterdam, account creation costs state gas beyond the 21000 limit, and the send ran out of gas.
• Full make fill; chain files copied into tests/.

Verification: make fill reports zero failures, make test (speccheck) passes with no known exceptions, make lint is clean.

Cross-client: full rpc-compat runs on 2026-08-25 against the glamsterdam-devnet-8 images (the newest devnet set with published images for all six clients), once on main's fixtures and once on this PR's, same images both times. 238 tests per client. The runs predate the 2026-08-28 update below; the fixture set was the same except for the five testing_commitBlockV1 tests, txpool_content, and estimate-simple-transfer.

client	main	this PR
go-ethereum	238	238
nethermind	238	179
erigon	238	169
besu	175	163
reth	209	155
ethrex	47	46

The new failures measure Glamsterdam implementation status, not fixture problems. A failing test means "differs from the geth-generated fixture", not "the client is wrong". Where they land:

• eth_simulateV1 is most of it (erigon 64, nethermind 57, reth 51, besu 11), concentrated in the identity fields of simulated blocks. The four clients that run it compute different blockAccessListHash values for the same simulated block. Nothing specifies the access list of a simulated block today; that needs an execution-apis decision.
• eth_estimateGas: geth returned 21000 for a simple transfer, nethermind 15000. Resolved: eth/gasestimator: return used gas for plain transfer estimates go-ethereum#35592 makes geth report the used gas, and the fixture now expects 15000. How estimates report EIP-7778 two-dimensional gas beyond the plain-transfer case is still open.
• testing_buildBlockV1: erigon and reth do not build Amsterdam blocks yet.
• eth_config: reth and nethermind differ from geth at an Amsterdam head.
• ethrex rejects the chain at block 1 (it supports post-merge networks only, Exit the client if on a pre merge fork lambdaclass/ethrex#5504); unchanged from main.

Known issue: hive's scheduled rpc-compat runs use release/master client builds. Against those, the post-Amsterdam tail of this chain fails until clients ship Glamsterdam, the same situation the Osaka bump created while Fusaka rolled out. Runs against the devnet images (table above) are the meaningful signal in the meantime. Note the devnet-8 geth image predates ethereum/go-ethereum#35592, so it fails estimate-simple-transfer against the current fixtures; geth built from master passes.

Update 2026-08-25: the table includes ethereum/hive#1585 (merged 2026-08-25), which defaults HIVE_TARGET_GAS_LIMIT to 60M in rpc-compat when the fixture env omits it. Verified by rerun: it fixes nethermind's two testing_buildBlockV1 divergences on both fixture sets (main is now fully green for nethermind), and changes nothing else. The erigon and reth testing_buildBlockV1 failures on this chain remain; they are Amsterdam block-building gaps, not gas-limit defaults.

Update 2026-08-28: rebased on main, which brought in the testing_commitBlockV1 spec and fixtures (#801). The commitBlockV1 generators now send slotNumber post-Amsterdam and their five fixtures are regenerated on this chain. go-ethereum bumped from the devnet-8 pin to master 86696a8f for the estimateGas and traceCall fixes above. Fill, speccheck, and lint all pass on the result.

[MysticRyuujin]

Related PRs, and what merging each would do to the failure table above.

Overlapping this PR

• fix: add slotNumber and targetGasLimit to testing_buildBlockV1 #862 adds slotNumber and targetGasLimit to the same testing_buildBlockV1 schema this PR touches, plus testgen wiring. Whichever lands second rebases: this PR carries the slotNumber half (sent as parent slot + 1, in both the buildBlockV1 and commitBlockV1 generators), fix: add slotNumber and targetGasLimit to testing_buildBlockV1 #862 adds targetGasLimit and the PayloadAttributesV4 title. Merging it does not change the erigon and reth testing_buildBlockV1 failures; those clients do not build Amsterdam blocks yet, whatever the attributes carry.
• Expose EIP-7843 slot number via JSON-RPC #806 specs the EIP-7843 slotNumber field on block objects. The fixtures here already carry it (eth_getBlockByNumber/get-latest returns "slotNumber": "0x3"), so Expose EIP-7843 slot number via JSON-RPC #806 gets fixture coverage from this chain as soon as it merges. It fixes no failures; the clients that import the chain already agree on the field.

Building on this PR

• debug: specify callTracer output and add debug_traceCall #855 (callTracer spec) is stacked directly on this branch.
• debug: specify callTracer output and add EIP-8037 two-dimensional gas to tracing #852 (EIP-8037 two-dimensional gas in tracing) gets post-Amsterdam traces to generate fixtures against once a client emits the fields. devnet-8 geth does not yet: the callTracer fixtures on this chain contain no regularGasUsed / stateGasUsed / gasRefund. Merging debug: specify callTracer output and add EIP-8037 two-dimensional gas to tracing #852 today changes nothing in the table and cannot be fixture-backed until geth implements it.

Merged context, exercised by this chain for the first time: #726 / #727 / #794 (EIP-7928 BAL methods and alignment) and #851 (BAL getter semantics). geth still returns -32601 for eth_getBlockAccessList and debug_getRawBlockAccessList, so the getters stay unfixtured.

Failures no open PR addresses

• The eth_simulateV1 wave (erigon 64, nethermind 57, reth 51, besu 11): the BAL and identity fields of a simulated block are specified nowhere. No open PR or issue covers it; it needs an execution-apis decision.
• eth_estimateGas post-Amsterdam (geth 21000, nethermind 15000) Resolved for plain transfers: eth/gasestimator: return used gas for plain transfer estimates go-ethereum#35592 (merged 2026-08-28) makes geth return the used gas, matching nethermind, and this PR's fixture now expects 15000. How estimates report EIP-7778 two-dimensional gas for calls that run code is still unspecified; debug: specify callTracer output and add EIP-8037 two-dimensional gas to tracing #852 covers tracing only.
• nethermind's base-fee validation of debug_traceCall: client behavior, to be filed on nethermind.

Adjacent but not Amsterdam-related: #742 moves the debug_getRawReceipts test to a post-Byzantium block, which would clear reth's pre-Byzantium receipt failures after a refill.

Update 2026-08-28: this PR now pins go-ethereum master 86696a8f instead of devnet-8 and is rebased on main, so it also carries the regenerated testing_commitBlockV1 fixtures from #801. Details in the PR description.