[flyq]

Summary

Bump mega-evm from f78e75c6 to 75743764, the revision mega-reth's sequencer payload builder pins since megaeth-labs/mega-reth#2553.
That revision adds MegaBlockExecutor::discard_transaction (megaeth-labs/mega-evm#405), which the sequencer calls after catching a per-transaction EVM panic, and makes every execution entry point's error path leave nothing in the journal.
The validator and the trace server still executed with f78e75c6, so the code that produces blocks and the code that validates them had drifted apart.

Running the same EVM revision as the sequencer keeps "any block the sequencer produces validates here" from resting on mega-evm#405's no-consensus-change claim alone.
It also lets mega-reth drop the second mega-evm copy from its lockfile once it repins stateless-*.

Why no call site changes

• mega-evm#405 states no consensus-visible change on any spec through REX6; the normal execution path is untouched.
• The validator, --block-list replay and the trace server build a fresh executor per block (per request for tracing) and never reuse one after an error or a panic: a panic fails the block through spawn_blocking's JoinError, and a trace aborts on the first failed transaction.
• None of the entry points whose error path #405 changed (inspect_tx, transact_one, system_call, …) is called directly; execution goes through MegaBlockExecutor::execute_transaction / run_transaction.
• The hardfork-derived BlockLimits (create_block_execution_env) already match mega-reth's replay_block_limits with no overrides: policy limits unbounded, consensus ceilings at protocol values, block gas limit from the header only.

Changes

• Cargo.toml: mega-evm rev f78e75c6 → 75743764.
• Cargo.lock: only mega-evm and mega-system-contracts move.
• bin/coverage-replayer/src/llvm.rs: example checkout path in a doc comment follows the new rev.

Note for coverage-replayer

coverage-replayer derives its measured scope and store namespace from the locked mega-evm rev, so stores scanned under f78e75c6 do not carry over; carry the candidate pool across with inspect --dump-pool before the next scan.

Tests

• cargo test --workspace: 622 passed, 0 failed (40 ignored by default, mostly the endpoint-backed debug-trace-server integration tests), including validate_block_mainnet_fixtures / validate_block_deriving_updates_mainnet_fixtures over all 63 mainnet fixture blocks and replays_mainnet_fixtures_with_header_sanity.
• cargo fmt --all --check, cargo sort --check, and both CI clippy configurations: clean.

🤖 Generated with Claude Code

[mega-maxwell]

Claude review status

Living comment — rewritten in place. The review workflow keeps this single comment up to date instead of posting a new one each round, so it always describes the latest reviewed commit and the earlier text is intentionally gone. No reply is needed here; reply to a finding in its own review thread, and answer an open question in a reply on this PR. The next review round reconciles your answer.

🛠️ Review did not finish

Attempted head 708c4979 · updated 2026-11-03T06:33:20+00:00

This round did not publish: MODEL_ACTION_FAILED: The Claude review retry action failed.. Anything listed below is from the last round that did. Workflow run: https://github.com/megaeth-labs/stateless-validator/actions/runs/41590540308