fix(tests,tools): make the EIP-7805 FOCIL tests fill on the current Amsterdam

[ilitteri]

Description

The FOCIL tests on this branch don't fill: 21 of 24 fail. Three independent causes, none of them in the tests' assertions. With these fixed all 24 fill, and the expected outcomes are unchanged.

1. t8n passes decoded transactions where raw ones are expected

check_inclusion_list_transactions takes Tuple[LegacyTransaction | Bytes, ...] and calls get_transaction_hash, which asserts isinstance(tx, (LegacyTransaction, Bytes)). The t8n caller passes convert_transaction(...), which returns a decoded fork transaction, so every typed-transaction case dies on AssertionError. Both tuples now go through encode_transaction.

2. The scenarios leave no gas budget for the block's own access list

They size block_gas_limit to their transactions alone. The EIP-7928 budget is block_gas_limit // GAS_BLOCK_ACCESS_LIST_ITEM, and the system-contract predeploys touched every block plus senders, recipients and coinbase come to ~30 items against a limit of ~22, so filling stops at BlockAccessListGasLimitExceededError.

Raising the limit alone would change what the scenarios assert, because the gas a pending inclusion-list transaction is allowed derives from block_gas_limit - gas_used_by_included_txs. So build_block raises the limit and spends exactly the same amount on ballast transactions, leaving remaining_gas — and therefore every "fits" / "does not fit" boundary — unchanged. The ballast is a whole number of empty transfers so the amount is always representable on the fork's calldata-cost lattice.

3. One funding transfer predates value-carrying transactions costing extra intrinsic gas

test_unsatisfied_when_block_tx_funds_pending_il_sender sizes and funds alice's transfer with calc(), but it carries value, so it needs calc(sends_value=True) and was failing INTRINSIC_GAS_BELOW_FLOOR_GAS_COST.

Assertions are untouched

Of the changed lines in test_focil.py, none touch expected_status, pytest.param, assertions, sender balances, nonces or inclusion-list composition — they are gas-limit plumbing, imports, and the sends_value fix. Scenario ids are unchanged, and the reference implementation validates all 24 expected outcomes during fill.

Deliberately not included

This branch's Amsterdam differs from glamsterdam-devnet-8 (ACCOUNT_WRITE 8000 vs 9000, CREATE_ACCESS derived from COLD_STORAGE_ACCESS rather than COLD_ACCOUNT_ACCESS), so a contract creation's intrinsic gas is 23000 here against devnet-8's 24000. That is a rebase question rather than a bug, so the gas schedule is left alone — flagging it because unsatisfied_with_contract_creating_pending_il_tx is the one scenario whose outcome moves with it.

Also worth noting: #3307 relocates evm_tools, so change 1 will need rebasing onto the new path if that lands first.

Testing

uv run fill tests/bogota/eip7805_focil --fork Bogota → 24 passed.

The filled fixtures were additionally run against a client implementation (ethrex) as a cross-check.

[jihoonsong]

cc @marioevz

[marioevz]

@ilitteri we need to rebase after #3307 lands anyway. I'll review this once we:

1. have merged 3307
2. rebased focil branch on top of the latest forks/amsterdam that contains 3307

[marioevz]

LGTM. I'll merge and then rebase on top of latest forks/amsterdam. Thanks!