Skip to content

kona,op-reth: extend TestForkActivationEVMRules with Karst and system-call contracts #24248

Open

Description

@sebastianst

Follow-up to #24193. Its TestForkActivationEVMRules activates a fork after genesis, runs probes in a block before and a block after the activation block, and proves those blocks with kona. #24193 restores the rules the deleted op-geth tests checked. Two more rules fit the same framework but go beyond that scope.

These are worth pinning because they live in code that kona and op-reth share (op-revm, alloy-op-evm, upstream revm and alloy-evm). A kona proof catches a divergence between kona and op-reth, but not a mistake both make.

Karst case

Karst has no case yet.

  • CLZ (EIP-7939): a status probe like the PUSH0 and TSTORE ones (PUSH1 1 CLZ) fails before Karst and succeeds after.
  • MODEXP input bound (EIP-7823): a call to 0x05 with a base longer than 1024 bytes succeeds before Karst and halts after. kona doesn't accelerate MODEXP, so both sides run the same revm code.

Karst's other Osaka changes (MODEXP and P256VERIFY repricing) change gas, not status or output, so they would need a probe that checks gas. Leave them out unless that is cheap to add.

System-call contracts at their activation block

In both cases the system call at the start of each block (alloy-op-evm) finds no code in the activation block, so it must do nothing there and start storing from the next block. That ordering is specific to OP chains; on L1 both contracts exist before their fork. A call through the Invoker in the after block, reading the parent beacon root or the parent block hash, would pin it.

🤖 Co-created with Claude Opus 5.5

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions