Skip to content

HF121-R3-REVIEW.md

Latest commit

 

History

History
78 lines (55 loc) · 3.55 KB

HF121-R3-REVIEW.md

File metadata and controls

78 lines (55 loc) · 3.55 KB
Raw

HF121 reviewed source revision r3

Review result

The original, unreleased HF121 source candidate required revision before production use. Revision r3 additionally isolates snapshot self-tests from the live root-owned publication lock. The consensus delta inherited from HF120 was correct and remains untouched, but the new operational surfaces needed memory, publication, and RPC hardening.

Load/memory review

  • status-fast now reads chain-status.json in normal operation.
  • A missing/stale cache triggers a serde streaming visitor that keeps only one block at a time and ignores persisted UTXO/mempool arrays without allocating them.
  • The snapshot publisher retains only the latest 4096 blocks while scanning.
  • The full canonical chain file is copied byte-for-byte instead of parsed and reserialized in memory.

Snapshot lock and self-test isolation

  • The live publisher keeps its default lock at /tmp/qub-mainnet-snapshot-publish.lock.
  • LOCK_FILE can be overridden for isolated tooling.
  • The self-test uses private lock files and logs inside its own temporary directory.
  • The self-test invokes the publisher through bash, so Windows archive extraction cannot break the test by dropping executable metadata.

Snapshot integrity review

Before publication, the publisher checks:

  1. Top-level network is mainnet.
  2. The blocks array exists and is complete.
  3. Every block references the exact previous block hash.
  4. Every block below #24000 is version 1.
  5. Every block at/above #24000 is version 2.
  6. Tail schemas and tip values are internally consistent.
  7. SHA-256 sidecars match the staged files.

tip.json is replaced last and serves as the generation commit marker.

RPC review

HF121 RPC is intentionally groundwork-only:

  • loopback only;
  • explicit rpc.enabled=true;
  • real auth token required;
  • token checked for every request;
  • no browser CORS;
  • bounded request headers and socket timeouts;
  • no enabled state-changing route;
  • template/submit routes return 501.

Consensus review

No HF121-r2 code changes the HF120 boundary:

height 23999 -> version 1
height 24000 -> version 2
height 24001 -> version 2

The source adds regression tests for this boundary and for fast-status metadata/stream fallback behavior.

Mandatory remaining gate

The archive has source-level parsing checks and the snapshot publisher self-test. A full project compile cannot be substituted by review: run cargo test, then release builds for both qubd and qub-core on the normal Windows build machine and on both seed droplets before public deployment.

Verification performed during the review

Completed in the review environment:

  • Rust syntax parsing with rustfmt across all src/ and tests/ files.
  • Shell syntax parsing for deployment scripts.
  • Snapshot publisher valid-generation self-test.
  • Snapshot publisher rejection test for an invalid post-activation block version.
  • Isolated Rust compilation/execution harness for the streaming status visitor and hardened state-file replacement.
  • Isolated Rust compilation/execution harness for loopback detection, RPC token comparison, and HTTP header parsing.
  • Mainnet/testnet/regtest config byte-for-byte comparison against HF120.
  • Consensus-anchor comparison confirming the #24000 version boundary is unchanged.
  • No-assets/runtime/secret file audit.

A complete QUB project build was not run in this review environment because the full pinned GUI/WGPU dependency registry is unavailable here. The normal cargo test and release build gates on the Windows build machine and both seed droplets remain mandatory.