We are not submitting ptc votes that we produce to our lcoal ptc op pool. So when we are the block producer we don't include our own ptc votes!
Co-Authored-By: Eitan Seri-Levi <eserilev@ucsc.edu>
Fixes a bug where a proposer votes payload missing on its own block.
The payload is published to the network but never imported locally. This PR adds gossip verification and import when a payload is sent to the http API
Co-Authored-By: Jimmy Chen <jchen.tc@gmail.com>
Closes:
- https://github.com/sigp/lighthouse/issues/8689
- Calculate the proposer index on the canonical chain (from canonical head) at `slot` and plumb it through to fork choice so it can be used to determine whether or not to apply the proposer boost. We use the proposer cache to handle state advances and avoid duplicate work.
- Update our FC tests to use `block.message().proposer_index()` (always pass), we are not attempting to test this feature in those tests. The EF tests use the correct canonical proposer idnex via `on_block`, except for invalid blocks which just auto-pass this check (these blocks get rejected by other checks in `on_block` anyway).
Co-Authored-By: Michael Sproul <michael@sigmaprime.io>
Payloads from the reprocess queue should be gossiped after import if they are still timely. In devnets this happens frequently since there are many cases where the envelope arrives before the block
Co-Authored-By: Eitan Seri-Levi <eserilev@ucsc.edu>
Implementation:
- payload_envelope_verification: implement the AvailabilityPending branch
in the envelope import flow. Previously returned
InternalError("Pending payload envelope not yet implemented") for any
envelope whose data columns hadn't yet been received, blocking the
end-to-end RPC import path. New `import_pending_execution_payload_envelope`
marks the payload as received in fork choice and persists the envelope to
the store; columns are still expected to arrive separately (gossip /
engineGetBlobs / reconstruction) and persist their own ops.
- sync manager: short-circuit `handle_unknown_parent_envelope` when the
parent's payload was received between gossip-verification and the trigger
reaching sync. No lookup is created; the trigger is treated as a no-op.
- gossip→sync hook: when a Gloas envelope is imported via the gossip path,
emit `SyncMessage::GossipEnvelopeImported { block_root }` so any lookups
awaiting that parent envelope unblock without depending on the in-flight
RPC response landing first. Closes the review-flagged race where a
gossip-imported envelope left child lookups pinned.
Tests (3 new):
- envelope_already_received_skips_lookup — trigger after envelope already
in fork choice creates zero lookups.
- happy_path_unknown_parent_envelope — end-to-end RPC import path: lookups
complete, head advances to the gossip block.
- happy_path_unknown_parent_envelope_via_gossip — pending envelope-only
lookup unblocked by a concurrent gossip import via the new sync hook.
Existing tests updated:
- bad_peer_envelope_rpc_failure / bad_peer_wrong_envelope_response now
expect the lookup to retry and succeed (mirroring `bad_peer_*` tests for
blocks/blobs/columns), reflecting the now-working import path.
- bad_peer_wrong_envelope_response — peer responds with a different block_root
than requested; request items raise UnrequestedBlockRoot, both lookups drop.
- crypto_on_fail_with_bad_envelope_signature — signature corruption rejected
in gossip verification, peer scored with lookup_envelope_processing_failure.
Rename the four already-landed tests to match the existing happy_path /
bad_peer / envelope_* / crypto_on_fail_with_* naming.
Wires the lookup test rig for Gloas:
- Capture per-block execution payload envelopes from the external harness
and serve them to peers via a new `network_envelopes_by_root` map.
- Handle `RequestType::PayloadEnvelopesByRoot` in `simulate_on_request` and
`Work::RpcPayloadEnvelope` in the simulator processor branch.
- Allow `TestRig` callers to override the genesis validator count and
bump initial balances to `max_effective_balance_electra` post-Electra,
which Gloas committee-selection requires for genesis init to converge.
Adds four tests for the parent-envelope-unknown flow (each verified
red/green by stubbing the corresponding source path):
- `creates_envelope_and_child_lookups` — `UnknownParentEnvelope` produces
exactly one envelope-only lookup for the parent root and one child
lookup awaiting that envelope.
- `idempotent_triggers` — repeated triggers for the same parent merge
into the existing envelope lookup; no duplicate lookups are created.
- `issues_payload_envelopes_by_root_rpc` — the envelope-only lookup
dispatches a `PayloadEnvelopesByRoot` RPC for the parent block_root.
- `drops_cascade_on_rpc_error` — when the envelope RPC errors, the
envelope lookup is dropped and the awaiting child cascades with it.
The end-to-end happy path (envelope arrives → child unblocks → block
imports → head advances) is gated on
`process_execution_payload_envelope` supporting `AvailabilityPending`,
which today returns `InternalError("Pending payload envelope not yet
implemented")`. That gap is independent of this PR's lookup machinery.
Replace `assert_event_journal_contains_at_least_ordered` helper with an
inline drain that just counts the gossip + reconstruction events. The
helper was carrying around `WORKER_FREED` bookkeeping and a strict
prefix-match for one caller; counting the two relevant work types until
both thresholds are met is the same check with much less code.
- block_verification: skip ParentEnvelopeUnknown check when parent is the
proto-array anchor. The anchor's `payload_received` is intentionally
false per spec (never added to `store.payloads`), but no envelope is
expected for it; without this exception the check rejects every
post-anchor gloas block.
- network tests: disable `engineGetBlobs` in the TestRig harness. Under
real crypto the mock EL's blob fetch raced the gossip path, importing
via a spawned task that the test didn't await -- leaving `head_root()`
unchanged when the assertion ran. The tests are designed to exercise
the gossip + data-column path; the engine fetch was incidental.
- network tests: relax `data_column_reconstruction_at_deadline` to allow
trailing duplicate reconstruction work items. The reprocess queue
removes its dedup entry on dispatch, so a column processed during an
in-flight reconstruction can dispatch a second one. The second is a
no-op via `reconstruction_started`, so accept >= 1 trailing event.
Store gossip-verified `PayloadAttestationMessage`s in the operation pool and pack them into the block body at during block production.
Built on top of #9145.
Co-Authored-By: Jimmy Chen <jchen.tc@gmail.com>
Currently, running `cargo check -p validator_manager` fails due to missing features. Although the `validator_manager` will almost always be called through the Lighthouse binary which will enable the required features, it is still good hygiene to ensure all workspace crates can compile standalone.
Add the `lighthouse` feature to the `eth2` dependency in `validator_manager`
Co-Authored-By: Mac L <mjladson@pm.me>
#6689
Inspired by the initial implementation of #9108, credit to @chong-he.
This adds an extension trait to `Hash256` and add a `short` method to provide smaller formatted hashes for logging.
Co-Authored-By: Mac L <mjladson@pm.me>
For gloas `attestation.data.index` should be set to 1 if we are attesting to a block whose slot is not the attestation duty slot and slot payload_status is `FULL`
Co-Authored-By: Eitan Seri- Levi <eserilev@gmail.com>
Co-Authored-By: Eitan Seri-Levi <eserilev@ucsc.edu>
Co-Authored-By: dapplion <35266934+dapplion@users.noreply.github.com>
We had a bug when performing batch builder exit signature verification. The EF spec tests cover this case, but the EF tests only calls individual signature verification (which is a separate code path). This PR unifies the two code paths. We should probably spend some time reviewing EF test code coverage and make sure we don't have separate code paths that do similar things.
Co-Authored-By: Eitan Seri-Levi <eserilev@ucsc.edu>
N/A
Do not make the AI agent always fmt and lint. This takes way too long and the agents I work with take this too literally sometimes and run lint after incomplete changes just wasting time.
I feel its not a big ask to run fmt and lint yourself and/or run it in some local configs instead of global ones.
Co-Authored-By: Pawan Dhananjay <pawandhananjay@gmail.com>
When producing a block at the fork, treat parent payload status as full
I've been testing on kurtosis and this fixes an issue where we cant propose a block at the fork.
This is a screenshot of the fix. The envelope shows missing because we are missing an SSE event, but the envelope is in fact being imported and the chain is progressing just fine
<img width="652" height="748" alt="image" src="https://github.com/user-attachments/assets/7764a68c-33fb-4987-a691-0af71f0bea02" />
Co-Authored-By: Eitan Seri-Levi <eserilev@ucsc.edu>
Just a little naming cleanup (no semantic changes) to remove mentions of pending and full states that were still lurking.
This hopefully helps Claude forget about the concept (it defaults to naming variables `pending_state`s without this change).
Co-Authored-By: Michael Sproul <michael@sigmaprime.io>
This reverts commit 2749e18d0e, from:
- #9092
We no longer need those changes since the abolition of pending/full states.
Co-Authored-By: Michael Sproul <michael@sigmaprime.io>
Fix a vulnerability in the slasher whereby it would OOM upon processing an invalid attestation with an artificially high `validator_index`. This fix has already been made available to affected users on the `slasher-fix` branch.
- Prevent attestations from being passed to the slasher prior to signature verification. This was unnecessary, as they would later be passed on successful validation as well.
- Add a defensive cap on the maximum validator index processable by the slasher. The cap is high enough that it shouldn't be reached for several years, and will quickly result in warning logs if forgotten.
- Add a regression test that confirms that the issue is fixed.
Co-Authored-By: Michael Sproul <michael@sigmaprime.io>
Fix a **consensus fault** in `PreEpochCache` 🙀
Fortunately it's only reachable on a network with `total_active_balance=0`, i.e. a network that's already completely dead. As such this PR is not time-sensitive in any way.
Add the floor on `total_effective_balance` when converting from `PreEpochCache` to `EpochCache`. An alternative would be to add the floor inside `PreEpochCache::get_total_active_balance`, however that would be redundant, as the only place this function is called outside this file is in single-pass epoch processing:
176cce585c/consensus/state_processing/src/per_epoch_processing/single_pass.rs (L461-L462)
The `set_total_active_balance` call already handles the floor.
A regression test is included.
Co-Authored-By: Michael Sproul <michael@sigmaprime.io>
N/A
Adds lints for rust 1.95. Mostly cosmetic.
1. .zip(a.into_iter()) -> .zip(a) . Also a few more places where into_iter is not required
2. replace sort_by with sort_by_key
3. move if statements inside match block.
4. use checked_div instead of if statements. I think this is debatable in terms of being better, happy to remove it if others also feel its unnecessary
Co-Authored-By: Pawan Dhananjay <pawandhananjay@gmail.com>
