Repository navigation
P8 field fix: SIEBCU first-connect retry, exact two-bit display and false no-report warning - #766
Conversation
…discovery association
…re with fresh client and evidence
…ts without guessing its semantics
… cancellation in bounded first-connect recovery
…eport warning after valid ingress
…d non-invented bit-string display
…neutral tone regressions
… measurement bits unchanged
CI red triage → targeted repair (2026-11-05)Failure captured: Build ARSAS #41966366626, commit
Actual corrections (no disabled tests):
New exact head: |
…wo-bit presentation Replace corrupted duplicate-tail formatter with baseline-preserving source and small pure presentation policy. Avoid numeric status inference on typed measurement/quality data. Add type-conflict regressions and preflight single-declaration invariant.
… and cached models Eliminate duplicate retry/teardown loops; preserve separate attempt budgets, enforce explicit cancellation, keep per-device scope, and classify only transient transport failures. Never replay discovery or MMS control writes. Add lifecycle guard.
Architecture consolidation after SIEBCU red-build audit (2026-11-05)Confirmed root of red builds, NOT an IED issueCI #41966366626 failed 5/3013 tests because a two-bit UI conversion escaped its declared status scope (4 legacy raw-Measurement regressions) and one stale source-string test required an obsolete timer log. CI #41967313612 then failed during C# parsing: Durable refactors committed, no vendor exception
Production safety gateCurrent exact head Next architectural stage should be measured, not speculative: benchmark association connect p50/p95 and first-report latency, per-association allocations on receive dispatch, bounded RCB overflow, and deterministic cancellation/reconnect fixtures in ARIEC P9 before promotion. ARIEC P9 #200, P8 #763, ARSAS main/RC, pinned baseline and stable v1.6.42 remain untouched by this PR. No claim of physical first-click recovery until operator evidence on a validated Windows trial. |
Cross-repo synchronization: P9 performance/read-mostly registry (2026-11-05)ARIEC61850 P9 PR #200 now at exact Consumer ARSAS PR #766 at |
P8 field-found SIEBCU regressions — safe consumer-only corrective trial
Base:
integration/reporting-engine-p8-20261029head37165a0cebd27ae9756810c540426f97b3bfad3a; do not modify P8 PR #763, ARSAS main/RC/stable, or pin a new engine.Actual operator evidence: 2026-11-05 SIEBCU at 192.16.1.192:102, initial connect logged 10:53:25 and second connect 10:53:35; second succeeded. Smart discovery 1.756s/211 requests; reporting 7/7 exact RCB streams, 7/7 selected values, 14 frame metadata. Nevertheless user-facing
USER2.LOCREMshowsbits(80, unused=6)and diagnostics mistakenly emit no Report within 3 s after seven streams had actual routed frames.Source-supported fixes
bits(80, unused=6)toBits [10]in operator-only formatter for non-DPC status, not invented Local/Remote. Position-DPC retainsClose [10]; malformed padding fails closed. OptionalBITtype marker only from explicit BitString metadata.Gates
Regression added for retry classification/association-only scope, four two-bit codewords, malformed values, DPC vs generic status, type marker and exact-RCB observer decision. CI must pass exact head. Keep PR Draft; a separate uniquely named Windows P8-derived EXE is required before declaring field acceptance. This is not P9 engine integration and does not prove SG7 #707 recovery.
Evidence:
docs/audits/P8_SIEBCU_FIRST_CONNECT_BITS_REPORT_20261105.md.