Test Fork Transition¶

Fork-transition tests for EIP-8038: State-access gas cost update.

"Same operation, different gas" across the Amsterdam boundary. A block at timestamp=14_999 runs under the pre-fork (parent) schedule; a block at timestamp=15_000 runs under the EIP-8038 schedule. Every before/after magnitude is derived from the opcode's own cost at each fork (bytecode.gas_cost / execution_cost / refund) — nothing is hardcoded.

Four proof styles are used:

• Opcode costs that are pure execution gas (BALANCE cold access, the EXT* code-read surcharge and the CREATE base) are measured exactly with CodeGasMeasure in each regime and asserted against the derived cost.
• Charges that cannot be measured from inside the frame that pays them (a value-bearing CALL, SELFDESTRUCT to a fresh beneficiary, the SSTORE clear refund) are pinned per block through the transaction receipt's cumulative_gas_used, which sums both gas dimensions and so pins a charge that the fork moves between them.
• Intrinsic repricings are pinned by setting gas_limit to the regime's exact minimum and pinning the receipt, and by straddling validity: one gas_limit that is sufficient before the fork and INTRINSIC_GAS_TOO_LOW after (access lists), or the reverse (authorizations, whose intrinsic falls).
• Model-level repricings the runtime cannot isolate (the SSTORE execution/state split) are compared across the two schedules via the bytecode's own cost methods while the operation is exercised in both blocks.