[Counting] Reviewing the counting for the L2 to L1 logs

[AlexandreBelling]

We recently found an issue in the counting logic for the L2L1 logs. It was invisible because the L2L1 logs are quite sparse. In the prover, we have a hard limit on how many L2L1 logs can happen at once in one conflated batch. The issue is about reviewing how this limit is taken into account by the conflator and the sequencer to ensure it is correctly enforced.

Definition

L2L1Logs counts the numbers logs happening in a conflated execution batch, such that:

• The topic[0] is the keccak-hash of the event signature MessageSent(address,address,uint256,uint256,uint256,bytes,bytes32) which is equal to 0xe856c2b8bd4eb0027ce32eeaf595c21b0b6b4644b326e5b7bd80a1cf8db72e6c
• The emitting address is the L2-bridge contract address. (0x971e727e956690b9957be6d51Ec16E73AcAC83A7 on sepolia)

Expected behaviour

• The sequencer should count the events satisfying the two above conditions and ensure it does not seal a block overflowing this limit.
• The conflator should ensure it does not create conflated execution batches that are overflowing the limit.

[letypequividelespoubelles]

Thanks @AlexandreBelling .

Could you tell the L2 bridge contract address for mainnet too ?

[letypequividelespoubelles]

To be sure, we have to count those LOG, whatever the data, and the topic[x] for x!=0 ?

[OlivierBBB]

I expect that given the emitting address is that of the L2-bridge contract which is our own there should be no room for "false alarms." Though I agree we want full confirmation.

[OlivierBBB]

There is some code that seems to implement a different logic @AlexandreBelling. From slack messages and this sentence in the issue

The sequencer should count the events

I gather that the issue is with the number of L2L1 logs (28 > 16) rather than the cumulative size of the Bytes message in the logs, which is what we seem to be counting at the moment. If so we may simply be missing a separate count for the number of these logs.

[OlivierBBB]

Also, given that the smart contract is ours counting the size's of the messages may prove redundant is say for instance if the logs are always constant size.