[masarray]

Stacked qualification milestone for #446, based on green P7.2 PR #466.

Goal

Turn the next physical retest into an explicit ingress-equivalence proof instead of relying on log interpretation.

Discovery and Open SCL are two bootstrap paths to the same IED. Once ARIEC has produced the configured-static acquisition plan, ARSAS now records a source-neutral semantic fingerprint and compares both ingress paths on the same logical IED.

Fingerprint authority

The semantic fingerprint includes:

• requested/catalog-mapped/uncovered counts;
• Static BRCB/URCB classification;
• exact DataSet identity;
• full ordered DataSet member identity + FC;
• selected runtime signal identity + FC.

Concrete live RCB slot names such as Rpt_ind01 vs Rpt_ind02 are excluded from the parity fingerprint because association ownership can legitimately select another indexed instance. Concrete slots are retained separately as runtime diagnostics.

Runtime behavior

• first Static DataSet run from either ingress records evidence and reports AWAITING PEER INGRESS;
• running the other ingress on the same logical IED compares fingerprints:
  • MATCH = source-neutral static semantics converge;
  • MISMATCH = explicit Discovery-only/Open-SCL-only semantic lines are logged;
• reconnecting the same ingress replaces that ingress evidence rather than creating false multi-session drift;
• no reporting/RCB activation behavior is changed.

Diagnostics

Copy Diagnostic now includes:

• Discovery semantic fingerprint;
• Open-SCL semantic fingerprint;
• parity status;
• concrete live RCB targets for each run;
• bounded semantic differences when mismatched.

Emergency diagnostics also preserve the parity summary.

Regression

Tests prove:

• Rpt_ind01 -> Rpt_ind02 does not change semantic fingerprint;
• ordered DataSet member drift does change the fingerprint and reports a mismatch;
• one logical IED can retain both ingress evidences;
• runtime/diagnostic integration does not use concrete RCB instance identity as the semantic parity key.

Physical 7SX85 retest remains required. Expected sequence: run Static DataSet from Discovery, then open the same SCD and run Static DataSet again; Diagnostics should report MATCH while allowing a different free indexed RCB slot.

[masarray]

P7.3 follow-up audit (exact head e95cbd68c0bc21bc1c85c29861a1cd3f656bce72): initial CI #3543/#569 failed on one overbroad source regression guard; 1,337 other tests passed. The guard was repaired to check the actual hash boundary rather than prohibiting legitimate concrete RCB names in diagnostic-only runtimeTargets. Additional fail-closed parity checks now require authoritative planning, selected points, nonempty static plans, complete ordered DataSet membership and selected bindings before an equal hash can produce MATCH. New INCOMPLETE EVIDENCE diagnostics cover empty plans, missing datasets and stale successes following a failed re-plan. Core ARIEC61850/RCB/GI/no-MMS-polling behavior is unchanged. Four commits applied on the same P7.3 branch. Fresh CI: Build #3546 and Field Capture #572 (pending at this comment). Physical 7SX85 Discovery→Open SCD proof still required; do not promote the CodeVerified engine trial or merge this stacked branch to main based on CI alone.

[masarray]

Additional exact-head audit: 389a6353756cb30f5f6550e4649a00391dc89242. The previously green-looking ordered-member-drift test was not a genuine SCL vs Discovery comparison because its Open-SCL fixture had no SclWorkspace model; corrected that fixture. Also repaired inherited .github/workflows/scl-interoperability-r7.yml: the branch had a cut-off PowerShell engine-provenance block plus ~269 duplicated malformed YAML lines, causing 0-job workflow failures. The workflow is now scheduled as real SCL Interoperability R7 Build #270 and retains explicit R7/R10 physical baseline + trial CodeVerified-only provenance checks. Exact-head Build #3549, Field Capture #575 and R7 #270 are still active as of this comment. No engine change, no mainline promotion; physical device parity test remains mandatory.

[masarray]

D17 evidence-gated historical reconciliation against protected integrated RC 817c3910e8223fa2ac531a4426ada5d2d4bbbe50 (no new merge to main).

• Current PR HEAD verified unchanged 389a6353756cb30f5f6550e4649a00391dc89242. GitHub compare HEAD→RC: behind=0, so the full historical commit ancestry already exists in the canonical RC. All 7 original changed paths remain present.
• Original regression suite tests/ARSAS.Tests/StaticAcquisitionIngressParityTests.cs (7 tests, 37 assertions) has an identical Git blob SHA at PR HEAD and RC, not merely a matching filename; historical positive/negative assertions therefore remain intact. 1/7 changed blobs match byte-for-byte; other production sources have evolved and must not be reported byte-identical.
• Source-level semantic mapping of evolved files: Canonical StaticAcquisitionParity models/tracker remain in RC: normalized DataSet member order/FC and semantic fingerprint, incomparable/missing-plans fail closed, RCB live slot excluded from semantic parity, and per-ingress Record/Compare. Evolved workflow and runtime/diagnostics preserved by newer source; physical traffic gate remains separate.
• Existing exact-RC-SHA Windows build, SCL R7 and reference guard are SUCCESS. No workflow has been rerun for this metadata-only reconciliation. Physical IEC 61850/BRCB/GI FAT, independent #520, overarching #446, and release gates are NOT certified by this closure.

Decision: close historical stacked PR as integrated into RC ancestry / superseded as an independent merge unit. Preserve full GitHub discussion and original immutable branch. Never cherry-pick again, do not mutate RC, stable v1.6.42, ARIEC engine pin, #753–#756 or parallel GUI/Write/SG7 lanes.