Open
Conversation
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
e6788aa to
93cc1fd
Compare
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
26f539f to
ec9aa2c
Compare
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
5c97886 to
6bd1123
Compare
domiwei
force-pushed
the
feature/caplin_gloas
branch
2 times, most recently
from
457288b to
ead33e8
Compare
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
826100f to
e8029ed
Compare
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
3b3cea8 to
1bae0e0
Compare
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
1bae0e0 to
29ba4f9
Compare
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
18b0a57 to
37c12f7
Compare
Emit SSE events for three new EIP-7732 gossip topics: payload_attestation_message, execution_payload_bid, execution_payload_available. Also fix PayloadStatus enum to match spec: Pending=0, Full=1, Empty=2. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…d sync fixes, and P2P hardening GLOAS envelope fetching: add by-range fallback after by-root failures, return full blocks from determineFullGloasBlocks for slot-based requests. Forward sync: add 30s request timeout, advance past empty slot gaps, update P2P status during sync, and set head state on stale handoff so ChainTipSync can receive gossip blocks. P2P fixes: separate transport errors from fork mismatches in handshake validation (keep peer on stream reset, disconnect on wrong fork), fix StatusV2 to include EarliestAvailableSlot (92 bytes), fix StateVersionByForkDigest to error on unknown digest instead of silently returning FuluVersion. Also: improve error messages in httpreqresp and rpc, add peer disconnect trace log, remove devnet-only workarounds (HTTP finality fetch, multi-peer retry, bootnode skip, metadata V2 fallback, gossip relaxation during sync). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
0e0249c to
7a293b8
Compare
…st data bug) The consensus-specs v1.7.0-alpha.4 proposer_boost and proposer_boost_is_first_block test data uses tick=51 for GLOAS, but the GLOAS BPS timing refactor lowered the attestation threshold to 3000ms (ATTESTATION_DUE_BPS_GLOAS=2500). At tick=51 the block arrives exactly at 3000ms into the slot, and 3000 < 3000 = false (not timely), so the proposer boost is not applied. The tick should be 50 for GLOAS. Our implementation matches the spec; the test data was not regenerated for the new timing. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…n routing (V5/V6) Add NewPayloadV5 and GetPayloadV6 to the EngineAPI interface and CL-side RPC client, fixing two bugs in the CL→EL version routing for GLOAS: - NewPayload for GloasVersion fell through to error default instead of V5 - GetAssembledBlock for GloasVersion was caught by >= FuluVersion routing V5 instead of V6 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ounds hardening - Add bounds checks in IsActiveBuilder, CanBuilderCoverBid, and InitiateBuilderExit before calling builders.Get() to prevent remote DoS panic from malicious builder_index values. - Guard ExecutionRequests parsing with version check to prevent nil deref in GLOAS block production (requests live in envelope). - Skip MEV-Boost blinded block path for GLOAS (Blinded() returns ErrGloasCannotBlind); builders use ePBS gossip bids instead. - Populate SignedExecutionPayloadBid for GLOAS self-build with correct spec fields (BuilderIndexSelfBuild, InfiniteSignature, slot, parentBlockHash, parentBlockRoot, prevRandao). - Route blob KZG commitments into the bid for GLOAS blocks. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement the "defer payload processing to next block" change from consensus-specs PR #5094 (merged Apr 16, 2026). This eliminates the dual-state model (pre-envelope / post-envelope) in favor of a single canonical post-state per block root. Key changes: Types & SSZ: - Add ParentExecutionRequests field to BeaconBody (GLOAS-only) - Add ExecutionRequestsRoot to ExecutionPayloadBid - Remove StateRoot from ExecutionPayloadEnvelope - Swap latest_block_hash / latest_execution_payload_bid positions (specs#5113) - Fix PayloadStatus constant values to match spec ordering (EMPTY=0, FULL=1, PENDING=2) State Transition: - New ProcessParentExecutionPayload (called first in ProcessBlock, before header) - New ApplyParentExecutionPayload (extracted helper for verify/apply split) - ProcessExecutionPayloadEnvelope becomes verification-only (no state mutations) - process_withdrawals uses inline FULL/EMPTY check - Epoch-aware builder payment with old-epoch direct withdrawal fallback Fork Choice: - Remove GetFullStateAtBlockRoot / GetExecutionPayloadState (single state model) - Simplify on_block parent state selection (no FULL state branching) - on_execution_payload: no state copy, verification only, stores envelope - Invalidate head payload status cache after envelope persistence - Add is_payload_verified (HasEnvelope) guard in ShouldExtendPayload - Add HasEnvelope check for PTC attestations in validate_on_attestation Block Production: - Use HasEnvelope && ShouldExtendPayload per spec's prepare_execution_payload - FULL path: copy state, apply parent execution payload, compute withdrawals - EMPTY path: use cached payload_expected_withdrawals - Hard error if head is FULL but envelope missing Forward Sync: - Determine FULL/EMPTY from consecutive block bids (count+1 lookahead) - Only request envelopes for FULL blocks via RequestEnvelopesFrantically - Remove anchor block from fork graph (not needed in single-state model) Cleanup: - Remove FetchFinalizedBlock from checkpoint sync - Remove IsParentBlockFull dead code - Remove findMissingEnvelopeRoots from chain_tip_sync Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…cs, and spec deviations Squash of fix/gloas-review-items addressing PR #18956 review findings: - Fix pendingCond.Wait() deadlock on context cancellation in gossip services - Return ErrIgnore (not nil) for already-seen data column sidecars - Fix SSZ EncodingSizeSSZ +4 offsets for variable-length ePBS types - Fix DecodeSSZ nil pointer deref via beaconCfg propagation - Fix getPayloadStatusTiebreaker to match EIP-7732 spec - Fix data race in OnPayloadAttestationMessage (GetState alwaysCopy=true) - Filter unsolicited envelope responses by requested root set - Return ErrIgnore when block not found for pending envelopes - Deep copy BlobKzgCommitments in ExecutionPayloadBid.Copy() Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
domiwei
added a commit
that referenced
this pull request
Apr 18, 2026
…cs, and spec deviations Squash of fix/gloas-review-items addressing PR #18956 review findings: - Fix pendingCond.Wait() deadlock on context cancellation in gossip services - Return ErrIgnore (not nil) for already-seen data column sidecars - Fix SSZ EncodingSizeSSZ +4 offsets for variable-length ePBS types - Fix DecodeSSZ nil pointer deref via beaconCfg propagation - Fix getPayloadStatusTiebreaker to match EIP-7732 spec - Fix data race in OnPayloadAttestationMessage (GetState alwaysCopy=true) - Filter unsolicited envelope responses by requested root set - Return ErrIgnore when block not found for pending envelopes - Deep copy BlobKzgCommitments in ExecutionPayloadBid.Copy() Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Squash merge of fix/gloas-spec-compliance branch (6 commits): Spec alignment: - Add block_access_list (EIP-7928) and slot_number (EIP-7843) to ExecutionPayload - Remove non-spec Slot field from ExecutionPayloadEnvelope - Add missing genesis checks in GLOAS state transition - Enforce ByteListSSZ max length and fix M1 parent_bid.slot - Add block.slot == payload.slot_number validation (EIP-7843) Audit fixes: - historical_states_reader: GLOAS version guard to prevent nil panic - block_production: use GetLatestBlockHash() for GLOAS builder parentHash - clstages: skip ExecutionPayload for GLOAS genesis blocks - spectests: use GetBlobKzgCommitments() accessor for GLOAS compatibility Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Persist and reconstruct all 9 GLOAS BeaconState fields so that ReadHistoricalState produces correct block roots for GLOAS-era slots. Storage strategies: - SlotData: latestBlockHash, nextWithdrawalBuilderIndex - Per-slot compressed SSZ: latestExecutionPayloadBid, builderPendingPayments, executionPayloadAvailability, ptcWindow - Dump+diff queues: builders, builderPendingWithdrawals, payloadExpectedWithdrawals Also includes: - DB schema version 7 migration - Snapshot table registration - ErrMissingGloasData sentinel with antiquary backoff - readCompressedSSZ passes actual state version (not hardcoded 0) - ReadRequiredQueueSSZ inlined to avoid double dump-table read - Comments on NewEth1Block constructors re: empty BlockAccessList - Comprehensive roundtrip tests Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…spectests Upgrade spectest fixtures from v1.7.0-alpha.4 to v1.7.0-alpha.5 and resolve all test failures across 8 forks (phase0 through gloas). Key changes: - Fix Eth1Header SSZ schema: remove incorrect GLOAS gate that added block_access_list_root and slot_number to ExecutionPayloadHeader (spec only includes them on ExecutionPayload). Fixes 10 LightClient "bad encoding size" failures. - Fix ExtraData nil panics: use NewEth1Header() in Clone/DecodeSSZ/New and add nil-receiver guards on ExtraData SSZ methods. - Fix envelope validation: use ValidatingMachine (BLS enabled) in applyEnvelope, return error for unknown beacon block roots. - Handle execution_valid=false: read execution.yaml metadata to skip CL-only tests where EL rejection cannot be simulated. Fixes 23 failures across bellatrix-fulu. - Cache SHA256 per i/16 group in ComputeBalanceWeightedSelection and computeProposerIndexElectra, reducing hash calls from 268M to 16.8M. GLOAS historical_accumulator test: 62s → 14s. - Implement gossip_attester_slashing and gossip_proposer_slashing spectest handlers with full three-state validation (valid/reject/ignore). - Register missing spectest handlers: execution_payload, parent_execution_payload, on_execution_payload_envelope, get_parent_payload_status, and several ssz_static types. - Un-skip proposer_boost fork choice tests (alpha.5 fixed tick values). - Fix RlpHeader for GLOAS: populate BlockAccessListHash (EIP-7928) and SlotNumber (EIP-7843) to prevent hash mismatches in NewPayload, sync, and block production. - Fix ListObjectSSZRoot deadlock in concurrent SSZ hashing. Results: 8 forks, 0 failures, 1 fixture-limitation skip (wrong_withdrawals). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…on, and spec compliance Squash merge of fix/caplin-gloas-audit-fixes into feature/caplin_gloas. Fixes: - cl/phase1/core/state/raw: fix baseOffsetSSZ GLOAS constant (2741117 → 3134333) and refactor EncodingSizeSSZ to stop double-counting fixed-size fields already in the base offset. Properly scope inactivityScores (Altair+), historicalSummaries (Capella+), and latestExecutionPayloadHeader (Bellatrix–Fulu only). - cl/sentinel/handlers: fix envelope handler tests that used a hardcoded startSlot=100, far below the minimum serve epoch derived from wall-clock time. - cl/beacon/handler: populate ExecutionRequestsRoot in self-build bid by decoding the EL's requestsBundle and computing hash_tree_root(ExecutionRequests). - cl/beacon/handler: broadcast self-built execution payload envelope on the execution_payload gossip topic so blocks transition from PENDING to FULL. Uses an LRU cache to bridge payload data between produceBeaconBody and broadcastBlock (envelope needs the BeaconBlockRoot from the final block). - cl/beacon/handler: select highest external builder bid from epbsPool when it exceeds local execution value, with proper ExecutionValue update in the API response. - cl/beacon/handler: proper envelope signing via validator client following beacon-APIs PR #580 pattern. Returns unsigned envelope in the v3 block production response; accepts signed envelope in the v2 block publish request. Falls back to InfiniteSignature with warning when no signed envelope provided. - cl/das: update stale TODO on GLOAS inclusion proof — consensus-specs v1.7.0-alpha.5 removes KzgCommitmentsInclusionProof from DataColumnSidecar; commitments verified via builder bid instead. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add missing GLOAS block to dynamic baseOffsetSSZ() (latestBlockHash, builders, builderPendingPayments, ptcWindow, etc.) — verified to match the hardcoded 3134333 for mainnet preset - Restore peerDas nil check in OnBlock EL blob availability path - Use config-aware MaxBlobCommittmentsPerBlock in NewBeaconBody - Add missing cltypes import in cache_accessors.go - Remove unused solid import in clstages.go - Remove duplicate PeerDas initialization in run.go Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Restore PayloadStatusNone case in on_block.go switch (defence-in-depth against EL returning unknown status string via Engine API) - Restore peerDas nil guard for Fulu data availability check to prevent nil pointer dereference when peerDas is not initialized - Fix proposerLookahead shallow copy in BeaconState.CopyInto to use CopyTo (prevents Merkle tree data race between copied states) - Gate forward sync stale timeout by fork version: 5min for pre-GLOAS chains (more conservative), 2min for GLOAS (handles new error paths) - Remove devnet debug logging from inserters.go (WriteTd read-back verification on hot path, hardcoded block range 3220-3250 log) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
BuilderPendingPayment.EncodingSizeSSZ() returned 8 when Withdrawal was nil, but the type declares Static() == true — its SSZ size must be constant (44 = 8 weight + 36 withdrawal). The nil guard caused baseOffsetSSZ() to under-count by 2304 bytes (64 elements × 36 byte shortfall), breaking GLOAS BeaconState SSZ encode/decode. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
2f8e480 to
3ba600d
Compare
NewForkGraphDisk persists the filled-in header Root back to the anchor state to prevent transitionSlot from recomputing HashSSZ. This is only needed for GLOAS where BlockRoot() uses the stored Root directly; for pre-GLOAS versions BlockRoot() always re-derives via HashSSZ, so the mutation makes the second BlockRoot call (in NewForkChoiceStore) return a different anchor root, breaking TestForkChoiceChainBellatrix. Gate the SetLatestBlockHeader call on Version >= GloasVersion. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
domiwei
force-pushed
the
feature/caplin_gloas
branch
from
aaf5552 to
bb8d2d3
Compare
…l panic NewEth1Block left Extra, Transactions, and Withdrawals nil, but getSchema() references them during HashSSZ(). This caused a nil pointer dereference when queuePendingEnvelope tried to hash an envelope with a freshly-constructed Eth1Block. Initialize these fields to match NewEth1BlockFromExecutionHeader. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The previous commit eagerly initialized Extra, Transactions, and Withdrawals in NewEth1Block, which changed JSON serialization from null to empty values and broke beacon handler golden file tests (TestHarnessPhase0 blocks). Move the nil guards into a shared ensureSSZFields() called by HashSSZ and EncodeSSZ. This prevents the nil pointer panic in queuePendingEnvelope without altering the JSON representation of freshly constructed blocks. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…coring Fix two EIP-7732 spec compliance issues found during audit: 1. Return HTTP 400 (not 500) when requesting blinded blocks for GLOAS-era blocks, since the blinded block concept does not apply post-GLOAS. 2. Add GossipSub topic scoring params for execution_payload_bid, payload_attestation_message, and proposer_preferences topics. Includes substring ordering fix (bid checked before payload to avoid false match). GLOAS weights excluded from maxScore() to avoid inflating penalties on pre-GLOAS forks. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…5125) When a GLOAS block has a pre-GLOAS (Fulu) parent at the fork boundary, getParentPayloadStatus accessed GetSignedExecutionPayloadBid() on the parent block which returns nil for pre-GLOAS blocks. The nil check then returned PayloadStatusEmpty instead of the correct PayloadStatusPending. Add a version check early — after getting the parent block but before accessing GetSignedExecutionPayloadBid(). If the parent block's version is less than GloasVersion, return PayloadStatusPending. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The spec simplified apply_parent_execution_payload by removing the parent_bid parameter. The function now derives it internally from state via s.GetLatestExecutionPayloadBid(). Update the interface in machine.go, the implementation in operations.go, and all call sites (ProcessParentExecutionPayload and block_production.go). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…compliance Address verified issues from PR #18956 review and Beacon API spec audit: - forkchoice: eliminate TOCTOU race in OnBlock pending envelope processing by extracting applyEnvelopeLocked and calling it while f.mu is held, deferring DB index writes to after unlock to avoid deadlock - fork_graph: add bounds checking in readBeaconStateFromDisk and ReadEnvelopeFromDisk to prevent panic on corrupt file lengths, with automatic buffer growth when needed - beacon/handler: add dependent_root and execution_optimistic to PTC duties response, matching attester/proposer duty patterns - beacon/handler: change POST pool/payload_attestations to accept array input with per-item error tracking, matching existing pool endpoint semantics - beacon/handler: add WithOptimistic and WithFinalized metadata to GET execution_payload_envelope response - beaconevents: fix execution_payload_available SSE event to emit flat {slot, block_root} per spec instead of full envelope with version wrapper Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…d panic Fix three issues discovered via Kurtosis local testing with Caplin CL + Erigon EL in GLOAS (EIP-7732) mode: 1. Block production sent zero head hash to ForkchoiceUpdated at the GLOAS fork boundary. The initial bid created by UpgradeToGloas has ParentBlockHash=0x0 (spec-compliant), but the EMPTY path in produceBeaconBody used this zero value as the EL head. Add fork boundary detection (Slot==0 && ParentBlockHash==0x0) to use parentBid.BlockHash instead, matching the spec's intent that pre-GLOAS blocks always had their payloads executed. 2. GetPayloadV1-V6 panicked on empty payload ID (index out of range [7] with length 0). Add decodePayloadID helper that validates the 8-byte length before calling binary.BigEndian.Uint64. 3. Block production continued to call GetAssembledBlock with an empty payload ID when ForkchoiceUpdated returned nil PayloadId (EL syncing). Add early return when idBytes is empty. 4. Gossip topic re-subscription at fork boundary: subscribeUpcomingTopics aborted on ErrExpiryInThePast for externally-managed topics (attestation, sync_committee, data_column_sidecar), preventing all subsequent topics from being created with the new fork digest. Also add missing RegisterTopicValidator for new fork digest topics. Also add gloas-caplin.io Kurtosis config for local GLOAS testing with Caplin. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The standalone caplin binary parsed --sentinel.bootnodes and --sentinel.staticpeers flags via SentinelCliCfg but never passed them to CaplinConfig, so the P2P manager never received bootstrap nodes. Also add gloas-caplin-mixed.io Kurtosis config for testing GLOAS with a mixed Lighthouse + Caplin setup. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
goCheckForkAndResubscribe ran its first fork check immediately on goroutine start, before RegisterGossipServices had populated any topics. This caused subscribeUpcomingTopics to iterate an empty list, then update forkDigest to the upcoming value — permanently missing the resubscription window. Move the ticker wait to the top of the loop so the first check happens after one slot tick (~6s), giving RegisterGossipServices time to complete. Also upgrade the fork digest change log from Debug to Info for observability, and log errors from subscribeUpcomingTopics. Also update gloas-caplin.io to use GLOAS-compatible Lighthouse VC image (bal-devnet-3) and single-node config. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.