Itzamna (2242c6ea) at 18 Mar 02:43

Summary

• rpcclient: Add 4 missing endpoint methods (GetLatestBlock, GetTransactionBlock, EstimateFee, PubkeyToAddress) to the HTTP client and interface
• testdata: Add privnet node config (genesis files, keys, Dockerfile), CI test-runner Dockerfile, and era fixture capture tooling
• tests: Add integration tests (sync, backfill, all 12 RPC endpoints, full tx lifecycle), preprod nightly tests, and CBOR era fixture deserialization tests
• CI: Add integration stage with privnet service, preprod scheduled job, and auto-build of test images on testdata/ changes
• docs: Document integration test setup, preprod tests, fixture capture, and CI pipeline stages
• Makefile: Centralize build/test/lint/docker commands as single source of truth for local development

Backfill Phase 2 fixes

Three fixes for the backfill block processing pipeline that was hanging with processed_blocks stuck at 0:

1. Remove WithDelayProtocolStart from backfill connections

In full-duplex NtN mode, gouroboros creates DemuxerChannels for all server protocols even with DelayProtocolStart. Without Start(), no goroutine reads from them — after 10 buffered messages the muxer blocks, halting ALL incoming TCP data including BlockFetch responses.

Replace manual Start() calls with auto-start + GetCurrentTip()/Sync() so the node treats workers as synced peers and serves BlockFetch requests.

2. Preserve rollback point and restore minSlot on resume

The Ouroboros ChainSync rollback point is the last common ancestor between forks — it must be preserved, not deleted.

• Change in-memory prune from slot < to slot <= so the rollback point stays in the buffer
• Change DB rollback queries from slot >= ? to slot > ? so the rollback point row is kept
• Restore minSlot from the highest processed point on resume, preventing the forward sync from picking up backfill headers and causing muxer deadlocks

3. SQLite write performance for backfill Phase 2

Phase 1 inserts ~2.4M headers, bloating the WAL to 200+ MB. The query planner chose idx_point_height for UPDATE queries, which degrades to a full table scan when most rows have height = -1.

• Use INDEXED BY idx_point_hash to force the correct index
• Add WAL checkpoints on startup, between phases, and periodically during Phase 2
• Reduce batch size from 10,000 to 1,000 for faster transactions
• Add _busy_timeout=5000 and reduce wal_autocheckpoint to 1,000 pages

Verification

• Phase 1 (header collection): ~25 min for 2.39M headers
• Phase 2 (block processing): ~2.2 hours, peak throughput ~93K blocks/min
• Forward sync: tracks node tip concurrently with zero errors
• All tests pass with -race

Commits

Bitol Maya (2bf18d46) at 18 Mar 02:10

... and 1 more commit

Bitol Maya (04cf15be) at 18 Mar 02:10

Official Radix Foundation seed nodes are all unreachable (port 30000 closed, DNS removed) as part of their 2026 infrastructure transition. Added StakeSafe community seed node as primary, keeping official ones as fallback

Official Radix Foundation seed nodes are all unreachable (port 30000 closed, DNS removed) as part of their 2026 infrastructure transition. Added StakeSafe community seed node as primary, keeping official ones as fallback

Itzamna (909e6c8d) at 18 Mar 02:03

Official Radix Foundation seed nodes are all unreachable (port 30000 closed, DNS removed) as part of their 2026 infrastructure transition. Added StakeSafe community seed node as primary, keeping official ones as fallback

Itzamna (2bd32d3e) at 18 Mar 01:22

... and 1 more commit

Itzamna (f18b1b16) at 18 Mar 00:57

Itzamna (f35bb68f) at 18 Mar 00:55

... and 14 more commits

Official Radix Foundation seed nodes are all unreachable (port 30000 closed, DNS removed) as part of their 2026 infrastructure transition. Added StakeSafe community seed node as primary, keeping official ones as fallback

Bitol Maya (7492a96c) at 17 Mar 18:28

... and 3 more commits

Bitol Maya (94a477c4) at 17 Mar 18:22

Bitol Maya (c6f57888) at 17 Mar 18:22

• Test/lint pipeline when directory is edited in MR
• Deploy pipeline when ts directory is edited so it auto deploys to npmjs when it's merged

Summary by CodeRabbit

• New Features

  • Pure JavaScript library added with JS transaction builder, signer, RPC helpers, address utilities, and new JS integration & e2e tests.

• Breaking Changes

  • Package renamed to @mayaprotocol/zcash-js.
  • Public API signatures and types updated (address/UTXO/transaction shapes and function names changed).

• Build & Infrastructure

  • JS build pipeline added (webpack, TypeScript config, ESLint, npm scripts) and CI jobs for JS testing and publishing.

Summary

• Score streaming swaps by deposit / remaining_sub_swaps instead of full deposit in scoreMsgs(), preventing large streaming swaps from monopolizing block space
• Freeze current swap queue logic as V128 and gate new behavior at version 1.129.0

Closes #259

Changes

1. Freeze V128: Copy current swap queue as manager_swap_queue_v128.go with renamed types
2. Version gate: Add 1.129.0 entry in GetSwapQueue() switch in managers.go
3. Core fix: In scoreMsgs(), create a local sourceCoin for streaming swaps that divides the deposit by remaining sub-swaps (Quantity - Count). Uses persisted streaming swap from keeper when available (accurate Count), falls back to msg-based swap for new entries (Count=0)
4. Tests: TestScoreMsgsStreamingSwap verifies fee-based scoring for streaming vs non-streaming, persisted progress, and Quantity=1 edge case

Key design decisions

• sourceCoin is a new local variable — does NOT mutate item.msg.Tx.Coins[0] (execution amounts unaffected)
• remaining > 0 guard prevents division by zero for completed swaps
• Error from GetStreamingSwap falls back to msg-based swap (conservative scoring)

Test plan

• TestSwapQueueVCURSuite — all 5 tests pass (4 existing + 1 new)
• TestSwapQueueV128Suite — all 4 frozen tests pass
• go build ./x/mayachain/... — clean build

Summary by CodeRabbit

• Bug Fixes
  • Improved SwapQueue operation handling with version-specific optimizations to ensure compatibility across different protocol versions.

        deadline := time.Now().Add(5 * time.Minute)
        for {
            if time.Now().After(deadline) {
                log.Fatal().Msg("timed out waiting for cardano to be ready")
            }
            res, err := http.Get(chainRPCs[common.ADAChain] + "/status")
            if err == nil && res.StatusCode == 200 {
                break
            }
            log.Info().Msg("waiting for cardano to be ready")
            time.Sleep(time.Second)
        }

Let's add a timeout for this. Not sure if 5 minutes it's enough.

Itzamna (a37f2527) at 17 Mar 17:16

Summary

• Score streaming swaps by deposit / remaining_sub_swaps instead of full deposit in scoreMsgs(), preventing large streaming swaps from monopolizing block space
• Freeze current swap queue logic as V128 and gate new behavior at version 1.129.0

Closes #259

Changes

1. Freeze V128: Copy current swap queue as manager_swap_queue_v128.go with renamed types
2. Version gate: Add 1.129.0 entry in GetSwapQueue() switch in managers.go
3. Core fix: In scoreMsgs(), create a local sourceCoin for streaming swaps that divides the deposit by remaining sub-swaps (Quantity - Count). Uses persisted streaming swap from keeper when available (accurate Count), falls back to msg-based swap for new entries (Count=0)
4. Tests: TestScoreMsgsStreamingSwap verifies fee-based scoring for streaming vs non-streaming, persisted progress, and Quantity=1 edge case

Key design decisions

• sourceCoin is a new local variable — does NOT mutate item.msg.Tx.Coins[0] (execution amounts unaffected)
• remaining > 0 guard prevents division by zero for completed swaps
• Error from GetStreamingSwap falls back to msg-based swap (conservative scoring)

Test plan

• TestSwapQueueVCURSuite — all 5 tests pass (4 existing + 1 new)
• TestSwapQueueV128Suite — all 4 frozen tests pass
• go build ./x/mayachain/... — clean build

Summary by CodeRabbit

• Bug Fixes
  • Improved SwapQueue operation handling with version-specific optimizations to ensure compatibility across different protocol versions.