Agenda
EOF Implementers Call #63 · Issue #1205 · ethereum/pm · GitHub
Notes
Recording
Meeting Notes
-
Clients and Compilers
- Solidity
- some functionality merged,
- updating solidity tests (proves correct compilation)
- working on stability next
- Frangio is working on a new EVM assembler for EOF and Legacy
- Do any assemblers output EOF as well? Other than YUL?
- EXCHANGE opcode textual representation, how? [n, m-n] or [n, m]?
- (mixed prefs, closer to bytecode/Internal encoding vs user meaning)
- proposal will be made for review in 2 weeks
- Solidity
-
Spec Changes
- Metadata
- Possibly change data to 0xff (and reserve 0x04)
- Not all containers use metadata in runtime containers
- initcode only,
- tx data only
- Event during initcode
- Keeping initcode separate from runtime code makes sourcify’s job harder
- metadata in tx only requires archive data of all deployment transactions
- There is an appeal to unreadable code in runtime, but perhaps we need different conventions for compiler metadata
- Mixing metadata with code produces runtime issues with indexes into the data and makes metadata discovery difficult.
- Generic Contracts and Deterministic / Counterfactual Deployments
- There is a desire for contracts at the same address on multiple chains
- Ways it’s done
- Nick’s method is used to create a single-use TX to make a fixed address (chain #0, high fees)
- CREATE2 deployer deployed w/ Nicks Method (most used factory)
- createx - GitHub - pcaversaccio/createx: Factory smart contract to make easier and safer usage of the `CREATE` and `CREATE2` EVM opcodes as well as of `CREATE3`-based (i.e. without an initcode factor) contract creations.
- createx factory is not implementable in EOF right now.
- Will TXCREATE fix this? (eof/spec/eof_future_upgrades.md at main · ipsilon/eof · GitHub)
- App devs use both salted create2 and nick’s method
- Nick’s method is hard on some other chains, no toehold.
- Can only deploy code in containers, not arbitrary contracts
- General question on how to get reliable addresses in Eth w/o “toehold process.”
- Unsalted deployed - nonce derived address
- EOFCREATE and TXCREATE require a salt, no nonce.
- Could the factory maintain this?
- Possibly remove the “witness” from the address. (`EOFCREATE`: Don't hash the init-container · Issue #162 · ipsilon/eof · GitHub)
- Witnesses include: initcode, code address (sender or caller), initcode container index
- Removing the witness makes counterfactual addresses more difficult.
- Time boxed at 15 minutes
- Metadata
-
Code Size EIP
- Needs to be accompanied by EOF changes to support > 64KiB containers / code / etc.
-
Testing
- Devnet 5 is the focus, no EOF work due to priority
- Fuzzing and Tracing overview
1 Like