dohertj2/mxaccessgw
1
0
Fork 0
Code Issues Pull Requests 2 Actions 12 Packages Projects Releases Wiki Activity

docs(archreview): mark WRK-06/11/12 + IPC-15 Done (P2 worker hot-path, windev-verified)

Browse Source 
Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
This commit is contained in:
Joseph Doherty
2026-07-09 16:00:29 -04:00
parent f61c816acf
commit c823a7b60b
1 changed files with 5 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
archreview/remediation/00-tracking.md
+5 -4
View File
@@ -92,13 +92,13 @@ Full design + implementation for each row lives in the linked domain doc under i
| WRK-03 | Medium | — | S | WRK-02 | Not started | Commands after shutdown starts are dropped with no reply |
| WRK-04 | Medium | — | M | — | Done | Envelope `sequence` can appear out of order on the wire |
| WRK-05 | Medium | — | M | — | Not started | One transient alarm-poll failure kills the whole session |
| WRK-06 | Medium | P2 | S | — | Not started | `MXSTATUS_PROXY` conversion reflects per field, per event |
| WRK-06 | Medium | P2 | S | — | Done | `MXSTATUS_PROXY` conversion reflects per field, per event |
| WRK-07 | Medium | P1 | M | WRK-04 | Done | Documented outbound write priority not implemented |
| WRK-08 | Low | — | S | — | Not started | Residual event queue discarded at graceful shutdown, undocumented |
| WRK-09 | Low | — | S | — | Not started | No `AppDomain.UnhandledException` hook |
| WRK-10 | Low | — | S | — | Not started | Top-level catch logs exception type but never message |
| WRK-11 | Low | P2 | S | — | Not started | Every accepted event is defensively cloned on enqueue |
| WRK-12 | Low | P2 | M | WRK-04 | Not started | No event batching per envelope; one flush per event; 25 ms poll |
| WRK-11 | Low | P2 | S | — | Done | Every accepted event is defensively cloned on enqueue |
| WRK-12 | Low | P2 | M | WRK-04 | Done | No event batching per envelope; one flush per event; 25 ms poll |
| WRK-13 | Low | — | S | — | Not started | Frame writer allocates a fresh buffer per frame; reader pools |
| WRK-14 | Low | — | M | — | Not started | Private-field naming split `_camelCase` vs `camelCase` |
| WRK-15 | Low | P2 | S | — | Done | Doc drift: STA thread name and heartbeat-counter note |
@@ -126,7 +126,7 @@ Full design + implementation for each row lives in the linked domain doc under i
| IPC-12 | Low | — | S | — | Not started | Gateway frame writer has no write lock; integrity rests on undocumented invariant |
| IPC-13 | Low | P2 | S | IPC-05 | Done | Gateway writer serializes twice and issues two stream writes |
| IPC-14 | Low | P2 | S | IPC-05 | Done | Gateway reader allocates fresh array per frame; worker rents from `ArrayPool` |
| IPC-15 | Low | P2 | M | — | Not started | Worker event delivery 25 ms poll with per-event write+flush, no batching |
| IPC-15 | Low | P2 | M | — | Done | Worker event delivery 25 ms poll with per-event write+flush, no batching |
| IPC-16 | Low | — | S | — | N/A | Correlation id carried twice per reply (Info) |
| IPC-17 | Low | P2 | S | — | Done | Two docs name the wrong Python generated-output directory |
| IPC-18 | Low | — | S | IPC-01 | N/A | Generated-code hygiene verified good — preserve under change (positive) |
@@ -253,6 +253,7 @@ Findings the review flagged as one coordinated design pass — sequence them tog
| Date | Change |
|---|---|
| 2026-07-09 | P2 Wave B — worker event hot-path (commit `f61c816`), **windev-verified**. WRK-06 → `Done`: `MxStatusProxyConverter` caches the four resolved `FieldInfo` per status type in a static `ConcurrentDictionary` (the `GetField` metadata scan ran 4× per status per event on the STA path); `GetValue`+`Convert.ToInt32` still run per event (late-bound RCW); both exception messages byte-identical (missing-field via `ResolveField`, not cached on throw through `GetOrAdd`; null-value unchanged). WRK-11 → `Done`: `MxAccessEventQueue.Enqueue` takes ownership of the passed `MxEvent` (stamps `WorkerSequence`/`WorkerTimestamp` in place, no `Clone()`); audited all 3 callers (base/alarm event sinks + provider-mode handler) — each builds a fresh event, none reuse it; `MxAccessValueCache.Set` now deep-copies its retained `Value`/`SourceTimestamp`/`Statuses` so the cache snapshot never aliases the queue-owned (later serialized) event. WRK-12 → `Done`: `WorkerFrameWriter` coalesces the flush across a drained batch (write each frame, one `FlushAsync` after the batch, then complete all) — preserves the written+flushed completion contract; a burst of N events costs 1 flush not N; on failure the in-flight batch + queue all fail so no caller hangs. IPC-15 → `Done`: the multi-event `WorkerEnvelope` body stays unimplemented (wire still one event per `worker_event` frame); gateway.md Performance section now distinguishes the shipped flush-coalescing from that deferred additive-proto change. **Windev:** x86 Worker builds clean (fixed a Windows-only `TreatWarningsAsErrors` CS8604 in the value-cache null-guard that macOS can't surface); Worker.Tests **356 passed / 0 failed / 11 skipped** (+4 new: converter cache-reuse, queue ownership-transfer, value-cache snapshot independence, writer batch-flush-once); gateway Tests **815 passed / 1 failed** — the 1 is the known windev-environmental SelfSigned-SAN assertion (passes on macOS), and all pipe-harness tests (which also exercise the earlier G-frame gateway frame change + WRK-12 end-to-end over a real pipe) pass. |
| 2026-07-09 | P2 Wave B — quick Mac items (commit `ec6f82b`). TST-11 → `Done`: `src/Directory.Build.props` single-sources the .NET-side version (Server/Worker/Contracts/tests stamped 0.1.2, was SDK-default 1.0.0) and appends the git short SHA to InformationalVersion (`0.1.2+<sha>`, guarded so a git-less build still works); verified Server assembly stamps `0.1.2+579282f`, Server+Tests build clean. Kept 0.1.2 (matches Contracts + aligned Python/Rust/Go); converging all to a single 0.2.0 cadence left as a release decision. WRK-15 → `Done` (docs-only path, no x86 build): STA thread-name refs corrected to the actual `MxGateway.Worker.STA` in `docs/WorkerSta.md` + `docs/MxAccessWorkerInstanceDesign.md`, and the stale heartbeat-counter note fixed (CaptureHeartbeat populates queue depth + sequence from the live queue). TST-23 → `Done`: already resolved by the Wave A gateway.md edit — the bidi `Session` RPC now sits under a "Future work: not implemented" heading (no "best long-term shape" framing), consistent with the proto. **TST-14 left open by design**: its only concrete step is deleting the user's untracked, gitignored `*-docs-*.md` working files (zero repo impact) — surfaced to the user rather than deleting files not created here. |
| 2026-07-09 | P2 Wave B — dashboard/security (commit `e15c3cb`). SEC-25 → `Done` (near-term hardening only; full per-session EventsHub ACL stays deferred to roadmap item 12 / TST-15): `DashboardEventBroadcaster` redacts tag values from a deep clone of the event before mirroring to SignalR when `Dashboard:ShowTagValues` is false (default), so the hub seam cannot leak values regardless of the missing ACL. Clears `MxEvent.value` (the OnDataChange/OnWriteComplete/OperationComplete/OnBufferedDataChange bodies are empty discriminators — values ride in the top-level field, incl. buffered arrays) + the alarm body `current_value`/`limit_value`; source event never mutated (shared with gRPC path + replay ring; verified). Makes the formerly-dead `ShowTagValues` flag live for the mirror. `EventsHub` TODO(per-session-acl) kept + tied to roadmap item 12. Test `DashboardEventBroadcasterTests` (3). SEC-30 → `Done`: `docs/Diagnostics.md` trimmed to mark opt-in command-value logging NOT YET IMPLEMENTED (no `LogCommandValues` knob, `RedactCommandValue` unwired, no values logged); wiring deferred pending secured-bulk redaction (SEC-13), not wired now. Server build clean; Dashboard tests 152/152. |
| 2026-07-09 | P2 Wave B — gateway event/frame hot-path performance (8 findings). **Frame I/O (commit `baae59c`):** GWC-08/IPC-13 gateway `WorkerFrameWriter` serializes once into one ArrayPool-rented buffer (LE length prefix + payload) and issues a single write (was: second serialization via `ToByteArray`, separate prefix array, two writes); IPC-14 `WorkerFrameReader` rents the payload buffer from ArrayPool instead of `new byte[]` per frame, read/parse length-bounded, returned in finally. Wire format byte-identical, cross-checked against the worker writer/reader. **Event/command hot path (commit `5e2e40a`):** GWC-06 `StreamEvents` timing via `Stopwatch.GetTimestamp()`/`GetElapsedTime()` (no per-event Stopwatch alloc); GWC-07/IPC-05 `MapEvent` transfers ownership of the inner `MxEvent` instead of cloning (safe — single-consumer `WorkerEvent`, MapEvent runs once pre-fan-out, all downstream consumers read-only; verified no post-mapping mutation) and `CreateCommandEnvelope` drops its redundant second clone (`MapCommand` already isolated the graph); every load-bearing isolation clone kept; GWC-15 `grpc_stream_queue.depth` converted from a per-event push counter to an `ObservableGauge` summing registered channel sources only at scrape time (name/semantics unchanged). **Replay ring (commit `cf1b3d4`):** GWC-14 replaces the `LinkedList` (node alloc per retained event) with a preallocated circular `ReplayEntry[]` (head+count), preserving ascending order, capacity eviction, time trim, oldest-read/ReplayGap math, both query paths, and the lock. → all `Done`. Server build clean (0 warnings); per-cluster tests green (WorkerFrameProtocol 10/10 incl. new near-cap round-trip, EventStream/Metrics/Distributor/Mapper 62/62, Distributor/Replay 33/33 incl. 2 new wrap-around cases). Full-suite macOS checkpoint: 769 passed / 44 failed, all 44 the known named-pipe/fake-worker-harness UDS-104-char limit (0 failures in any touched area). **Pending: windev run** to exercise the frame I/O over a real pipe (the 44 pipe-harness tests are macOS-blind). |