Inconsistency between deducted L1-fee and calculated L1-fee in receipts

[ezdac]

Currently we short-circuit the l1-fee calculation within the STF and the mempool so that the L1CostFunc that is used in the BlockContext or mempool environment just returns 0 without any considerations for the SystemConfig blobbasefeeScalar and basefeeScalar parameters:

op-geth/core/types/rollup_cost.go

Lines 127 to 129 in 48d7851

	if config.IsCel2(blockTime) {
	return func(rcd RollupCostData) (fee, gasUsed *big.Int) { return new(big.Int), new(big.Int) }
	}

In comparison to that, the l1 cost function that is used for calculating fields in transaction receipts does not include the short-circuit cost function. It thus returns e.g. Ecotone based values for the fees in the receipt:

op-geth/core/types/rollup_cost.go

Lines 263 to 273 in 48d7851

	// extractL1GasParams extracts the gas parameters necessary to compute gas costs from L1 block info
	func extractL1GasParams(config *params.ChainConfig, time uint64, data []byte) (gasParams, error) {
	// edge case: for the very first Ecotone block we still need to use the Bedrock
	// function. We detect this edge case by seeing if the function selector is the old one
	// If so, fall through to the pre-ecotone format
	// Both Ecotone and Fjord use the same function selector
	if config.IsEcotone(time) && len(data) >= 4 && !bytes.Equal(data[0:4], BedrockL1AttributesSelector) {
	p, err := extractL1GasParamsPostEcotone(data)
	if err != nil {
	return gasParams{}, err
	}

This can be verified by querying a random Alfajores transaction:

 cast rpc -r https://alfajores-forno.celo-testnet.org "eth_getTransactionReceipt" 0xb436b9b8c7eda68de4efeaa4907326fd3e207ce361e80d133d075472239716fd | jq
{
  ...
  "l1BaseFeeScalar": "0x558",
  "l1BlobBaseFee": "0x1",
  "l1BlobBaseFeeScalar": "0xc5fc5",
  "l1Fee": "0x304b58869",
  "l1GasPrice": "0x161075592",
  "l1GasUsed": "0x640",
 ...
}

This is not compatible and results in confusion from an end-user perspective.

We should set the l1 fees to zero via the SystemConfig scalar mechanism and remove the short circuit function.
That way, we achieve code-parity with upstream, a consistency between fee-deduction and receipt values and zero l1 fee.

Related: celo-org/optimism#225