You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Fix EVM spec_id for RPC calls: all EVM execution paths (eth_call,
eth_estimateGas, eth_createAccessList, debug tracing, bundle simulation)
now derive the correct SpecId from chain config + block header instead of
defaulting to PRAGUE.
Move serve.rs from signet-node to signet-rpc: transport infrastructure
(HTTP/WS/IPC) is now reth-free, using tokio::spawn and Handle::current()
instead of reth's TaskExecutor. This removes axum, tower-http, interprocess,
and ajj dependencies from signet-node.
Load StorageRpcConfig from host RpcServerArgs: rpc_gas_cap,
max_tracing_requests, and gas price oracle settings now flow through from
reth's CLI args instead of always using defaults.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
**Problem**: RPC reads can span hot and cold storage at different points in time. Reorgs can make resolved block numbers stale between hot and cold queries.
41
+
42
+
**Approach**:
43
+
- Document the consistency model in `StorageRpcCtx` rustdoc
44
+
- Verify all cold queries use explicitly-resolved block numbers (already mostly the pattern)
45
+
- Detect reorg-induced staleness in `get_filter_changes` (coupled with reorg tracking above)
46
+
- Consider hash-based verification: resolve number from hot, fetch header hash, pass hash to cold for consistency check
0 commit comments