[kyscott18]

Summary

Execute

• execute runs once, keeps that post-state, and records the touched accounts and slots as the journal's access list.
• execute's gas_used is the gas with cold access. No runtime code reads it.

Simulate

• simulate runs the discovery pass (skipped when the journals give a cached access list), then one pass at max gas with the access list warm. The binary search and the final pass are removed.
• gas_limit = gas_used + reserve. Each frame reserves 2,301 gas for the SSTORE sentry (EIP-2200), and enough that 63/64 of its remaining gas covers each child's gas plus the child's reserve (EIP-150). Precompiles get no SSTORE reserve. Monad has no refunds, so gas_used is the gas spent.
• At that limit, the tx takes the same path as at max gas. The limit can be up to about 2,300 gas above the true minimum.
• A halted call uses all the gas it is given, and its reserve counts that gas. So a caught halt in a call with a fixed gas budget gets a gas limit that works. A caught halt in a call that forwards all its gas would give a gas limit near u64::MAX, which the node rejects. Typewriter contracts never catch a halt in a nested call (a failed call reverts the tx), and harness.rs documents that the gas limit relies on this.

Other

• The access list leaves out all precompiles of the active spec. Before, the filter covered only 0x01–0x0a, so P-256 at 0x100 was in the access list.
• ExecuteResult and SimulateResult are separate types in Rust and TS. SimulateResult no longer has journal_id or slot_writes.
• harness.rs is simpler: one Journal type writes pre or post images, access lists merge through ordered sets, and dispatch uses one response helper.
• Every EVM run goes through run_tx, which always takes the run's state out of revm's journal (finalize), so no path can leave state behind. execute commits the state it gets back.
• The wire structs hold alloy types (Address, U256, U64, Bytes, B256, B512, revm's AccessList) and MonadHardfork, and serde does the hex. The hand-written parse and format helpers and the hex crate are gone. Slots and values go out as B256, so the wire format is unchanged. Bad params fail while parsing the request, so dispatch reads the request id on its own to answer with it.
• DB writes that cannot fail (Infallible) match Ok(()), journal writes take Image::Pre or Image::Post instead of a bool, and item comments are doc comments.
• The typewriter-evm test script now also runs cargo test --lib.

Interface

• createEVM(initParams) builds the harness with its init params. The init request, the initialized flag, and the "init already called" error are removed.
• slot_writes is { slot, value }[] and holds only writes to the called contract.
• keccak_preimages is Hex[], the 64-byte KECCAK256 inputs (a mapping key and the mapping's slot). The inspector no longer records 32-byte inputs or hashes them. Only execute returns them.
• execute and simulate no longer return output.
• Results carry status: "success" | "revert" | "halt" in place of success and revert_data. A revert has hex data, a halt has a reason (e.g. InvalidFEOpcode), and journal_id is only on a successful execute. Before, a halt came back as revert_data: "halt: <reason>", a string typed as Hex.
• typewriter adds MutationHaltedError and SettlementSimulationHaltedError, exported from the package root, so a halt is no longer reported as an unknown revert.
• typewriter.ts and db-query.ts use the new shapes, and enqueueMutation has a TODO: the enqueue's journal is never pruned.

Results

Same machine, PR head before these changes (9008e45) against 4bc8675:

	before	after
packages/typewriter/test/harness.benchmark.ts (3 runs)	3,765–3,775 mutations/s	3,694–3,847 mutations/s
typewriter-evm simulate, discovered, burn 5000	5.15 ms	3.30 ms
typewriter-evm simulate, cached, burn 5000	3.72 ms	1.43 ms

The harness benchmark shows no measurable change: simulate does not limit it. typewriter-evm execute throughput varied from 99k to 133k calls/s between runs, so it gives no clear result.

Test plan

• bun run lint, bun run typecheck, bun run test
• Rust tests: the tx fails at gas_used and succeeds at gas_limit for an SSTORE sentry, precompile calls (0x01, 0x100), and nested calls, and a caught halt in a call with a fixed gas budget.
• The access list leaves out 0x100 and matches eth_createAccessList.
• execute reports a halt (0xfe) as status: "halt", separate from a revert.
• A request with bad params is answered with its own id.
• createEVM fails with an unknown spec (serde's unknown-variant error). packages/typewriter tests pass against Postgres (27 pass), and the simulation test passes with SEED=0x01.

🤖 Generated with Claude Code

[greptile-apps]

Gas tests lose coverage

The gas tracker was replaced, but the tests for CREATE and calls that send value were deleted without replacements. The remaining call tests all send zero value, so mistakes in those gas rules could pass unnoticed. Keep tests that run these transactions at the returned gas_limit and check that they succeed. This is non-blocking test feedback.

Prompt To Fix With AI

This is a comment left during a code review.
Path: packages/typewriter-evm/src/harness.rs
Line: 897-900

Comment:
**Gas tests lose coverage**

The gas tracker was replaced, but the tests for `CREATE` and calls that send value were deleted without replacements. The remaining call tests all send zero value, so mistakes in those gas rules could pass unnoticed. Keep tests that run these transactions at the returned `gas_limit` and check that they succeed. This is non-blocking test feedback.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

[greptile-apps]

Returned gas limit can fail

simulate can report success but return a gas limit that makes the transaction revert. For example, a contract requiring gasleft() > 100_000 passes the run at u64::MAX, but a simple frame gets only 2,301 gas added to its measured cost. The returned limit then fails the same check.

The old search tested whether the transaction succeeded at the chosen limit. The new reserve only covers call forwarding and the SSTORE check. Keep a fallback that finds a working limit for contracts that check remaining gas.

Prompt To Fix With AI

This is a comment left during a code review.
Path: packages/typewriter-evm/src/harness.rs
Line: 534-535

Comment:
**Returned gas limit can fail**

`simulate` can report success but return a gas limit that makes the transaction revert. For example, a contract requiring `gasleft() > 100_000` passes the run at `u64::MAX`, but a simple frame gets only 2,301 gas added to its measured cost. The returned limit then fails the same check.

The old search tested whether the transaction succeeded at the chosen limit. The new reserve only covers call forwarding and the `SSTORE` check. Keep a fallback that finds a working limit for contracts that check remaining gas.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.