Clarification on EIP 3860

[davidmurdoch]

Looking into Shanghai for Ganache and in reading EIP 3860 it isn't clear to me if the raw transaction.data field should be subject to the limits of initcode, as transaction data isn't the same thing as initcode. Ethereumjs tx package does check the length of the raw transaction data:

ethereumjs-monorepo/packages/tx/src/eip2930Transaction.ts

Line 173 in 1fbc077

checkMaxInitCodeSize(this.common, this.data.length)

.

[davidmurdoch]

We've talked to the hyperledger besu team and they believe the new EIP 3860 initcode limit should only be considered during contract creation, and not used to limit the raw transaction data field limit.

https://discord.com/channels/905194001349627914/905205502940696607/1082835346829619241

Invite link: https://discord.gg/hyperledger

cc @jochem-brouwer @holgerd77

[jochem-brouwer]

What discord server is this? I cannot read this discussion or see what server it is.

It seems to me that you have indeed found a bug. To clarify: initcode max length check should only occur if there is a create transaction (to field is null / undefined / "empty"). However, if you send a "normal" transaction (i.e. to a contract with a big load of data) then this is not subject to this initcode size check.

[jochem-brouwer]

Regarding this (very-most-likely bug), see: #2575

[jochem-brouwer]

Feel free to contact me on Discord also :)

[davidmurdoch]

Do you know of a canonical source for an EIP-relative definition of initcode? https://www.rareskills.io/post/ethereum-contract-creation-code describes the data field for a contract creating transaction from, Solidity's perspective, as <init code> <runtime code> <constructor parameters>. And another blog, https://monokh.com/posts/ethereum-contract-creation-bytecode, describes it similarly.

On Stack Exchange https://ethereum.stackexchange.com/questions/76334/what-is-the-difference-between-bytecode-init-code-deployed-bytecode-creation describes initcode === contract creation code === data field.

I'm not taking a stance either way, I'm just trying to understand the terms.

[davidmurdoch]

I found an implementation of the initcode size check in geth, so if we want to treat their implementation as canonical, it would seem initcode === tx.data, with the empty to address check just as in your latest PR: https://github.com/ethereum/go-ethereum/blob/793f0f9ec860f6f51e0cec943a268c10863097c7/cmd/evm/internal/t8ntool/transaction.go#L174-L177

[jochem-brouwer]

Yes, this sounds/looks absolutely right to me.

Again, the gotcha here is: the fields initdata or data (or whatever you want to call it) of tx are the same fields. The to field determines the behavior:

• It is empty: create a contract
• It is not-empty: normal transaction

Please note: if you send a tx to the zero-address this is still a normal transaction, since to field is the zero address and not "non-zero".

Hence, by 3860, check iff (if-and-only-if) to field is empty (zero-length buffer/null/undefined), if data field does not exceed MAX_INITCODE_SIZE. If it does, tx is invalid and it is rejected over all nodes/pools, and will never be included in any block. (This is covered by #2575 )

[jochem-brouwer]

Side note: this should be covered by tests, and this should also be tested on a testnet.