fix(iota-core,iota-protocol-config): do not lock immutable objects in post-consensus conflict resolution (#12602)

[filipdulic]

Description of change

Under the P-COOL flow, post-consensus conflict resolution derives its owned-object lock set from the raw transaction bytes, where ImmOrOwnedMoveObject covers immutable objects too. An immutable object's reference never changes and the per-epoch lock table is insert-only, so the first transaction reading e.g. a CoinMetadata locks it for the rest of the epoch — every later reader is dropped with ObjectLockConflict.

The fix, gated behind a new pcool_skip_immutable_object_locks feature flag (rides the existing version-33 arm, which is unreleased — devnet still advertises 32; set for all chains, inert without enable_pcool_flow):

• The main path acquires locks from the object set the post-consensus deny checks already load, which excludes immutable inputs. No new reads.
• An already-executed transaction derives its lock set from its own effects (old_object_metadata): every consumed owned input appears there, immutable inputs never do, and effects are identical on every validator regardless of execution progress.
• The conflict probe stays on the raw byte-derived set — an immutable reference can never be found in any lock tier.
• handle_transaction_validation_checks now reports transaction-owned and authenticator-owned references separately. Existing callers (certificate-flow signing, soft locks) consume only the transaction set, so behavior outside P-COOL is unchanged. The authenticator list is provably empty today (validation only admits immutable or shared authenticator inputs); it exists so post-consensus locking keeps covering those inputs if that rule ever changes.

Links to any relevant issues

fixes #12602

How the change has been tested

• Basic tests (linting, compilation, formatting, unit/integration tests)
• Patch-specific tests (correctness, functionality coverage)
• I have added tests that prove my fix is effective or that my feature works
• I have checked that new and existing unit tests pass locally with my changes

Unit: the repro (two readers of one immutable object), a flag-off counterpart pinning the old behavior, and the already-executed branch — including a failed execution and an owned input the Move call never mutates. E2E: reading the IOTA CoinMetadata twice within one epoch, on both schedulers. Protocol-config snapshots regenerated (one line per chain).

Release Notes

• Protocol: Adds pcool_skip_immutable_object_locks (protocol version 33): immutable transaction inputs no longer acquire owned-object locks in post-consensus conflict resolution. Inert unless the P-COOL flow is enabled.
• Nodes (Validators and Full nodes): Fixes transactions being rejected with ObjectLockConflict against an immutable input (e.g. a CoinMetadata) for the remainder of an epoch under the P-COOL flow.
• Indexer:
• JSON-RPC:
• GraphQL:
• CLI:
• Rust SDK:
• gRPC:

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01QHZ8T9kqPKHyM6s7gDuSpr