EOF Implementers Call #6

[lightclient]

(FKA EOF Breakout Room)

Meeting Info

• Jan 11, 2023, 15:00 UTC
• Duration: 60 minutes
• Zoom: https://ethereumfoundation.zoom.us/j/85004720429?pwd=K1hnV1Z4bE1RZW50VEZTQnlCNHNuUT09
  📅 Subscribe to the Ethereum Protocol Call calendar for calendar invites

Agenda

• Client updates
• Compiler updates
• Spec updates
  • Remove PC verbiage
  • Support in tracing
  • Initcode argument support
  • Unified EIP?
• Testing updates
  • Ipsilon updates
  • Parsing and validation test format proposal - with filled tests
  • Fork format going forward - shanghai+eof or cancun or something else?
• Interop plans
  • Focus on EOF1 vs EOF2
  • Goals before interop
  • Goals after interop

[shemnon]

EIP-4750 EOF Functions

• Can we remove the notion of the start of the code section being PC 0? With the removal of the PC operation all PC is relative, and Besu has an optimization where we don't have to slice up the code in the EVM. Change the verbiage to something like this in CALLF

8. Sets `current_section_index` to `code_section_index` and `PC` to the first byte in the code section, and execution continues in the called section.

and remove , and PC is set to 0. from the Execution section. This aligns with making PC unobservable in EOF, clients would be free to set their own mappings.

EIP-3155 EVM trace specification interactions

• How do we reflect code section in the EIP-3115 format? (PC=0 from the prior section may impact this too, but it could be defined as PC relative to the start of the code section)
• Do we want to put the return stack items in the trace as an option (like memory)?

[rodiazet]

Current EIPTests report for evmone and geth. https://hackmd.io/@etgDMnCFQgS7tnZ9yjkOPw/Sk4RrJsco

[chfast]

Spec update: ethereum/EIPs#6261

[lightclient]

A few decisions from the call:

• Explain everything must be relative if PC=0 is removed.
• Section always for EOF code, PC relative.
• Change expected exception from integer to string.
• Use cancun instead shanghai+eof
• Need to discuss delegatecall
• Goals before interop
  • pass all released tests
  • implement reference test runner
  • think about spec changes that still need to be made for EOF1
  • bonus: think about EOF2 (banning most of code introspection) and consider how compatible EOF1 is with that view

[poojaranjan]

Recording - https://youtu.be/PKwMCmxHcas
Notes - https://hackmd.io/@poojaranjan/EOFBreakOutNotes#-Notes-6