Skip to content

🌊 Wave 2026-07-11 v3 · аудит 9 never-audited модулей — spike-sentinel глушит spike (P1) + flow_control credit underflow (P1) + quarantine 8-bit таймер (P2)#73

Draft
gHashTag wants to merge 1 commit into
mainfrom
feat/wave-report-2026-07-11-v3

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Draft. Только docs/WAVE_REPORT_2026-07-11-v3.md. Не мёрджить (docs/AUTONOMOUS.md). Закрывает-частично issue #72.

Разобраны 9 сгенерированных модулей (src/lib.rs:21-46), не смотренных прошлыми волнами, + specs/*.t27.

Честность (проверено, не по памяти)

Новые находки (спек-уровень, переживут фикс codegen)

  • W1 P1 anomaly_detector.t27:62,249,266TYPE_SPIKE=0 совпадает с sentinel «нет аномалии»; if (anomaly_type != 0) глушит все spike.
  • W2 P1 flow_control.t27:149,37,85,99,171MSG_CREDIT_UPDATE без клампа → window-credits u32-underflow → ложный backpressure + мусорный credit_grant (удалённый триггер).
  • W3 P2 quarantine_manager.t27:16,29,90 — 8-бит start_time vs QUARANTINE_DURATION=1000 → таймер релиза сломан/underflow.
  • W4 P2 adaptive_routing.t27:82,87,101255/x целочисленный score → сдвиг к path index 0.
  • W5 P2 multipath_routing.t27:154,164 — несогласованный модуль round-robin.

Heatmap, наука→предписание (Kung&Morris 1995, RFC 1982, ETX MobiCom 2003, NIST 3σ), план на 4 спринта, 3 линии кооперации, граница — в отчёте.

Граница

Audit + plan; без имплементации; без железа; без owner-gated crypto.

phi^2 + phi^-2 = 3

New spec-level findings surviving the (upstream-tracked) codegen fix:
- W1 P1 anomaly_detector.t27: TYPE_SPIKE=0 collides with "no anomaly"
  sentinel -> all spike anomalies silently dropped (:62,249,266)
- W2 P1 flow_control.t27: unclamped MSG_CREDIT_UPDATE -> window-credits
  u32 underflow -> false backpressure + garbage credit_grant (:149,85,99,171)
- W3 P2 quarantine_manager.t27: 8-bit start_time cannot time a 1000-tick
  quarantine -> release timer broken / underflow (:16,29,90)
- W4 P2 adaptive_routing.t27: 255/x integer score -> index-0 bias
- W5 P2 multipath_routing.t27: distribute_load %total_paths vs %4 mismatch

Codegen breakage (8/9 modules) NOT re-reported: tracked in PR #44,
issue #61, t27#1401/#1456/#1457. rustc 1.95.0 verified per-file.

phi^2 + phi^-2 = 3

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@gHashTag

Copy link
Copy Markdown
Owner Author

Triage — KEEP

Verdict from cascade analysis: KEEP, awaiting caskade merge.

Root cause of red CI (build + test (-sim baseline)): committed gen/rust/ on feature branches with expanded module set contains E0425 defects from three optimizer scan bugs in t27c (const_propagate, copy_propagate, dead_store_elim not recursing into control-flow bodies). Not a defect in this PR — this PR is docs-only and has no code paths affected.

Cascade to unblock (all three DRAFT):

  1. t27#1461 — mut-inference (base t27/master)
  2. t27#1462 — recursive optimizer scan (stacked on #1461), verified −182 errors cross-env (linux sandbox 28 / macOS arm64 26, ±2 epsilon)
  3. tri-net#80 — regen gen/rust/ on feat branches with new t27c

Once #80 lands, baseline CI on this PR goes green automatically. No action required on this PR — it stays DRAFT until human merge.

Realm-check discipline: numbers above measured on commit SHA bbaaa4a1 in t27, cargo check --lib grep count ^error\[E. Not fabricated.

phi^2 + phi^-2 = 3

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants