feat(relay): support relaying all bundles from a tx with --all

[vibelyova]

Problem

bundle relay only ever processed the first InteropBundleSent event in a transaction. Any transaction that emits multiple bundles (e.g. a batched operation sending to several destination chains) would have all bundles after the first one permanently stranded — with no error or warning.

The --msg-index flag existed to pick a specific bundle, but the bundle-selection code always selected the first InteropBundleSent regardless of the flag's value, so it had no real effect.

Solution

• Add --all flag to bundle relay. When set, every InteropBundleSent in the receipt is relayed in sequence (proof → root → execute/verify for each).
• Fix --msg-index to actually select the bundle with that index. If no bundle matches, a clear error suggests using --all.
• Extract find_bundle_log_positions — the core function that scans logs in order, counts L1MessageSent events from 0x8008, and derives the correct msg_index for each bundle. This matches the logic already used by auto-relay but was missing from the relay command.

Usage

Relay all bundles from a tx (new):

cast-interop bundle relay \
  --chain-src era --chain-dest test \
  --tx 0xSOURCE_TX_HASH \
  --all \
  --private-key $PRIVATE_KEY

Relay a specific bundle by index (fixed):

cast-interop bundle relay \
  --chain-src era --chain-dest test \
  --tx 0xSOURCE_TX_HASH \
  --msg-index 1 \
  --private-key $PRIVATE_KEY

When using --all with --out-dir, each bundle's artifacts are written with a _{msg_index} suffix (bundle_0.hex, proof_0.json, etc.) to avoid overwrites.

Tests

9 unit tests added for find_bundle_log_positions covering:

• Empty log list
• Single bundle
• Two sequential bundles with correct msg_index values
• Unrelated L1MessageSent events before the first bundle shifting indices
• L1MessageSent events from wrong addresses (not counted)
• Bundles from wrong InteropCenter address (not returned)
• Bundle appearing before any L1MessageSent (saturating_sub edge case)
• Mixed centers in same receipt

Additional fixes (code review follow-up)

Critical

• Fix panics on user input: root_wait called .unwrap() on parse_b256 from --expected-root (panics on bad hex); abi.rs called .unwrap() on U256→u64 conversion in decode_bundle_status/decode_call_status. Both now use ? error propagation.
• Fix tx show ignoring --center overrides: was hardcoded to INTEROP_CENTER_ADDRESS constant instead of using the resolved AddressBook, so config and CLI overrides had no effect.
• Fix relay not waiting for destination tx receipt: sent the handler tx and reported success without checking if it reverted on-chain. Now waits for receipt and bails on revert, matching send and token commands.

Important

• Fix bundle extract only extracting the first bundle: silently ignored multi-bundle transactions. Added --all and --msg-index flags reusing the same extract_bundles logic from relay.
• Fix dead require_signer_or_dry_run in send.rs: the check was called after the dry_run branch had already returned, making it unreachable. Moved before the branch so it actually validates.
• Add --private-key-env to auto-relay: was the only command without it, forcing users to pass keys on the command line (visible in ps and shell history).

Minor

• Fix debug println! corrupting --json output: tx_show printed "Decoding L1MessageSent event..." to stdout; bundle_action printed debug messages to stdout. Moved to eprintln or removed.
• Remove write_relay_outputs unnecessary async: function did only synchronous fs::write calls.
• Remove unnecessary calldata.clone() in relay dry-run branch.
• Remove unused INTEROP_CENTER_ADDRESS constant from types.rs.
• Fixed all pre-existing cargo clippy -- -D warnings errors (redundant closures, clone on Copy types, dead code branch in doctor.rs, import hygiene, etc.).

🤖 Generated with Claude Code