0gfoundation/0gchain-Aristotle v1.0.7

Aristotle-v1.0.7

on GitHub

Summary

This is an RPC-only release: it fixes eth_estimateGas correctness on 0G and removes a
CPU-amplification vector in the estimator. No consensus, state-transition or payload-building
changes — validators are not required to upgrade for consensus safety.

1. eth_estimateGas no longer tracks the request gas cap (critical)

The 0G revm fork raises gas_used to 80% of the gas limit and clears the refund at the end of
execution. The estimator seeded its optimistic candidate from gas_used + gas_refund, so the
candidate landed at ~80% of the cap and the returned estimate tracked the cap, not the actual
usage. Observable symptom: with a 36M cap, balanceOf on Bridged USDC
(0x1f3AA82227281cA364bFb3d253B0f1af1Da6473E) returned 29,029,739, while 0g-geth returned
34,013 for the identical request — a ~850× overestimate.

Estimates are now seeded from the top-level frame's pre-settlement gas, recorded by a
read-only inspector and raised to the EIP-7623 calldata floor, matching 0g-geth's MaxUsedGas
behaviour. The lower bound and the binary search are unchanged; the inspector only runs for the
first full execution.

Impact: wallets, dApps, bridges and gas oracles that call eth_estimateGas on 0G were receiving
gas limits roughly an order of magnitude too large.
Verification: re-run the request above against a 36M cap — expect ~34,013, not the cap.

2. eth_estimateGas is no longer a CPU amplification vector (hardening)

The EIP-7623 calldata floor was recomputed on every frame end — N × O(M) wasted work for N nested
frames and M bytes of calldata — and both N and M are caller-controlled through state overrides.
The floor is now computed once per transaction.

• 200k nested CALL frames + 128 KiB calldata: 9,566 ms → 73.7 ms (plain execution: 72.3 ms)
• 200k nested CALL frames + 4 KiB calldata: 371 ms → 74.0 ms

Testing

• Real-EVM eth_estimateGas regressions added (four of them fail on the previous estimator).
• 76-row case/cap request matrix (crates/rpc/rpc/testdata/gas_alignment.json) shared with a
  0g-geth reference test.
• cargo test -p reth-rpc --lib: 39/39 passed, including 8 estimate_gas_* regressions.

Update Priority

See Update Priorities for more information about this table.

All Changes

• perf(rpc): avoid redundant EIP-7623 floor recomputation per frame
• fix(rpc): seed estimate_gas candidate from pre-settlement gas