[masarray]

Field issue #493

Private user BCU_GE.iid verified 13 typed magnitude phase paths under WYE MMXU1.A (6) / PhV (4) and DEL PPV (3), plus top-level CMV AuxV.cVal.mag.f. The original private IID, endpoint and device metadata are not committed. This patch uses synthetic IEC schema for tests.

Root cause: 124/124 static memberships were present but top-level structured WYE/DEL FCDs had no unique scalar PrimaryValueReference; selection dropped the phase measurements even though the six existing RCBs routed genuine reports.

Changes:

• Iec61850StaticPhaseLeafProjection resolves only named FC=MX, typed WYE/DEL phase magnitude attributes through the existing schema-safe leaf resolver, with explicit per-phase q/t.
• Preserve the original mandatory whole-FCD membership; add exactly proven descendant phase value rows as separate runtime candidates, idempotently.
• Static report-only selection admits all exact typed leaf identities under the configured DataSet, without prefix guess or another RCB.
• Allow scalar CMV/MV top-level FCD (e.g. AuxV) to use its existing unique canonical magnitude. WYE/DEL root remains rejected as a single scalar.
• Synthetic schema regression covers six current channels, four phase voltage channels, three line-voltage channels, AuxV, repeated inventory merge and sparse/shuffled source schema.

Hard invariants: report-only, no cyclic MMS process polling, no dynamic DataSet write, no new commands, no control/reporting engine patch, engine pin unchanged 352c81e6a798635c6addcee0683235ca87ad416d.

Draft only until full Windows CI and field replay on GE F650 demonstrate report-origin values for all available phase leaves. Build/CI is not real RCB traffic evidence. Do not merge/release without physical acceptance.

[masarray]

Exact-head 2026-10-09 qualification checkpoint

HEAD 3cf821428efd954e415e7455acbd2936ff0019cd: 9/9 GitHub workflows SUCCESS, including Smart Discovery mainline/merge/provenance/budget/repeat-run/production guards, SV evidence validator, Windows native Build ARSAS, and Field Capture Build.

Build ARSAS run #37934966712: compilation SUCCESS, app regression tests SUCCESS (including new synthetic WYE/DEL whole-FCD phase tests), publish SUCCESS, portable single EXE smoke SUCCESS, artifact upload SUCCESS.

Exact-head artifacts:

• Win-x64 portable single EXE #11617806826.
• Field Capture Win-x64 #11617737183.

The original private BCU_GE.iid was used only for local verification of typed IEC 61850 phase names/CDC/scope; it was not uploaded to GitHub. Synthetic regression fixtures contain only generic IEC 61850 model shapes.

CodeCommitted/CIPassed/ArtifactAvailable: YES. Real F650 physical values/units/quality/ingress and static per-RCB continuity: PENDING. Keep #494 draft and do not merge or release before user validates phase currents, phase voltages, line voltages, AuxV, control/status, and report-only behavior.

[masarray]

Reconciliation D1 — evidence-backed disposition (2026-10-26)

Disposition: B — historical P7.6F implementation superseded by the canonical integrated RC; closing this tracking PR, not promoting the RC or certifying physical F650 interoperability.

Canonical replacement: ARSAS RC 817c3910e8223fa2ac531a4426ada5d2d4bbbe50 (protected draft integration #727/#729), retaining this PR head 3cf821428efd954e415e7455acbd2936ff0019cd as an ancestor: GitHub compare original HEAD → RC reports ahead 1908, behind 0, merge base = original HEAD. The original branch and commits remain untouched.

Unique delta: All five PR-touched paths verified against the current RC. Exact blob preservation for four:

• Services/Iec61850DataSetSignalInventoryService.cs = f3c956350577839c3d4adf4e83ef0c0ac97a88f3
• Services/Iec61850StaticPhaseLeafProjection.cs = 2b83b096e3a9de44b17e94bc24585ccbd0abb918
• Services/SchemaSafeAggregateProjectionService.cs = 65183fd23793b10230bdeb3b9c913a66c369e794
• tests/ARSAS.Tests/StaticWholeFcdPhaseMeasurementTests.cs = e5089993faa30ddb9c17fc32eaa28ca9cafc8084

The fifth service Services/Iec61850StaticDataSetAuthoritySelection.cs evolved from eb94e455f4265b773084396ed37f845a7bf2fd36 to 8e9eaf4e5cdb333675584f4c3756c91080f62b14. Current implementation builds one ToLookup(NormalizeLiteral(DataSetReference), OrdinalIgnoreCase) instead of re-filtering a snapshotted device.Signals catalog for every member. The membership predicates and all original controls/phase leaf/scalar candidates, exact object/display reference equality, unique phase match requirement, candidate prioritization, authoritative RCB backing and SCL-vs-live model policy remain in the current function. This is a read-only performance/indexing refactor, not a fuzzy matching or fabricated measurement fallback. The byte-identical two-test suite continues to assert exactly 13 WYE/DEL phase magnitudes, correctly typed q/t, one scalar AuxV, idempotence, missing-phase fail-closed, FC=CO rejection, and root DO scope.

Safety: No code, tag, binary, checksum, engine pin, branch or CI workflow mutated by this administrative closure. No claim of live-wire F650 FAT or R7/R10 success; existing #493/#495 field issues and #688 performance lane remain open. Reproducible regression test is preserved exactly in protected RC. Current-tree selection uses a one-pass lookup; if signal collection mutates concurrently during Build, that separately requires a concurrency audit rather than assuming field qualification.

Follow-up: Preserve #727/#729 release/physical gates; verify this canonical selection in multi-IED F650 FAT and runtime/performance qualification via #493/#495/#688. Reopen this PR if a specific uncovered regression is demonstrated. No Windows CI rerun warranted for the PR bookkeeping operation.