Claude finished @svlachakis's task in 3m 6s —— View job

Code Review — eip7778-gas-fix

• Read changed files and diff
• Analyze gas-accounting logic
• Verify test correctness
• Check for regressions and edge cases
• Post findings

Summary

The fix is correct and clean. The root cause was that transaction admission under EIP-7778 used _blockCumulativeReceiptGas (sum of SpentGas, i.e. post-refund) as the gas budget gate, while header.GasUsed was accumulated with EffectiveBlockGas = Math.Max(preRefundGas, floorGas) (pre-refund). This asymmetry allowed transactions to be admitted when the pre-refund block gas would exceed the limit, which is the bug.

The fix removes the now-unnecessary _blockCumulativeReceiptGas field and collapses the admission check to always use header.GasUsed directly. Because header.GasUsed already tracks the correct basis per-spec — pre-refund under EIP-7778, post-refund without it — no spec-switch is needed at the admission site.

Gas accounting path

Test analysis

The renamed test Transaction_admission_uses_pre_refund_block_gas_when_eip7778_enabled correctly inverts the assertions from the buggy behavior:

• tx1 creates a refund gap: blockGasAfterTx1 > receiptGasAfterTx1
• GasLimit is set to blockGasAfterTx1 + GasCostOf.Transaction - 1 — exactly 1 gas short for tx2
• Under the old code tx2 was admitted (using post-refund basis, so there appeared to be room)
• Under the new code tx2 is correctly rejected: admission uses header.GasUsed = blockGasAfterTx1, leaving only GasCostOf.Transaction - 1 available

One structural point worth noting: when _processor.Execute(tx2, new BlockExecutionContext(block.Header, ...), ...) is called, the extension method calls SetBlockExecutionContext internally, which resets _blockCumulativeRegularGas = 0. header.GasUsed lives on block.Header and is not reset, so the admission check still sees the pre-refund gas from tx1 — correct.

Findings

No Critical, High, or Medium issues. One low-level observation:

Low — The inline comment added at the admission site (TransactionProcessor.cs:680) correctly explains the EIP-7778 behavior, but omits that EIP-8037 takes a separate early-return path. This is very minor since the EIP-8037 guard directly above it makes this clear to a careful reader. No action required.

Verdict: approved — no blocking issues. The fix is minimal, correct, removes dead code, and is backed by a tight regression test.

Claude finished @svlachakis's task in 2m 27s —— View job

Code Review — eip7778-gas-fix

• Read changed files and diff
• Analyze gas-accounting logic
• Verify test correctness
• Check for regressions and edge cases
• Post findings

Summary

The fix is correct and clean. The root cause was that transaction admission under EIP-7778 used _blockCumulativeReceiptGas (sum of SpentGas, i.e. post-refund) as the gas budget gate, while header.GasUsed was accumulated with EffectiveBlockGas = Math.Max(preRefundGas, floorGas) (pre-refund). This asymmetry allowed transactions to be admitted when the pre-refund block gas would exceed the limit — the bug.

The fix removes the now-unnecessary _blockCumulativeReceiptGas field and simplifies the admission check to always use header.GasUsed directly. Because header.GasUsed already tracks the correct basis per-spec — pre-refund under EIP-7778, post-refund without it — no spec-switch is needed at the admission site.

Gas accounting path

Test analysis

Transaction_admission_uses_pre_refund_block_gas_when_eip7778_enabled correctly inverts the assertions from the buggy behavior:

• tx1 creates a refund gap: blockGasAfterTx1 > receiptGasAfterTx1
• GasLimit is set to blockGasAfterTx1 + GasCostOf.Transaction - 1 — exactly 1 gas short for tx2
• Under the old code tx2 was admitted (post-refund basis created apparent headroom)
• Under the fixed code tx2 is correctly rejected: gasUsedForAllowance = header.GasUsed = blockGasAfterTx1, leaving only GasCostOf.Transaction - 1 available

One structural subtlety: _processor.Execute(tx2, new BlockExecutionContext(block.Header, ...), NullTxTracer.Instance) calls SetBlockExecutionContext internally (via ITransactionProcessor.cs:77), which resets _blockCumulativeRegularGas = 0. However header.GasUsed lives on block.Header and is not reset by SetBlockExecutionContext, so the admission check correctly sees the pre-refund gas from tx1. Since tx2 is rejected before the Complete: label, header.GasUsed is never modified — the third assertion block.Header.GasUsed == blockGasAfterTx1 holds.

Moving tracer.EndBlockTrace() to before tx2 is correct — tx2 is executed outside the block tracer session.

Findings

No Critical, High, or Medium issues.

Low — The inline comment at TransactionProcessor.cs:680 correctly explains the EIP-7778 basis alignment, but omits that EIP-8037 takes a separate early-return path above it. This is very minor since the guard is immediately visible above. No action required.

Verdict: approved — no blocking issues. The fix is minimal, correct, removes dead code, and is backed by a tight regression test.

Need to also raise as test gap for EEST test so consensus is enforced on the change https://github.com/ethereum/execution-specs/issues

That not failing with the behaviour change confirms a problem

Test is here ethereum/execution-specs#2932