[masarray]

Scope: real code, one atomic commit, no redundant CI

Baseline ARSAS #504 exact HEAD f90f4697fb1ee607bb5500119f6b5298ce01879f; source includes cumulative #501 GOOSE/NTP/Discovery improvements. Engine #157 and consumer immutable pin unchanged.

Root fix: the original LooksLikeLoopback name-substring filter falsely excluded Microsoft's KM-TEST virtual Ethernet adapter, while the fallback selected the last usable physical NIC without evidence. This branch adds a pure and replay-testable GooseAdapterSelectionPolicy:

• Same-PC MMS target IP ownership takes precedence over the unicast route; multiple IPv4 aliases on one Windows NIC remain the same identity.
• Npcap/Windows exact GUID is preferred, then a unique valid noncontradictory MAC; never guess from friendly-name substrings or a sole unrelated adapter.
• Real Windows KM-TEST NIC is eligible, but dedicated Npcap NPF_Loopback pseudo-interface remains excluded from automatic Ethernet GOOSE selection (manual exploration still possible).
• IED-card click starts read-only observation only on unique identity match; otherwise preserves one-shot manual selection/start. Status/tooltip clearly distinguishes MMS route evidence from actually observed Ethernet GOOSE packets.
• Retains existing Npcap transport, BPF, cancellable capture, bounded UI queue and static report-only/control authority. No native code, extra scan, active publisher control, dynamic DataSet writes, timer or per-packet GUI loop.

Tests: synthetic Windows NIC/Npcap snapshots: KM-TEST same-host, conflicting/unavailable GUID/MAC, multiple aliases vs duplicate address owners, remote route, pseudo-loopback, duplicate capture adapters and UI state contracts. No raw customer IID/capture included.

Explicit boundary: This is P3A identity selection, not a proven GOOSE publisher probe or field acceptance. Need local same-PC publisher verification with actual 0x88B8 frames, optional P3B bounded probe only if field evidence demands it. P2 C264 typed FCD reporting is separate and remains open. Run only naturally-triggered exact-head Windows CI; do not manually rerun/push successive minor commits. Keep DRAFT and leave physical mainline guard in place.

[masarray]

Exact-head qualification — 2026-10-10

HEAD: 5e55fa2022bf3f79f1d747a176e75365306bdcd9, exact parent ARSAS #504 f90f4697fb1ee607bb5500119f6b5298ce01879f (compare: 1 ahead, 0 behind, 6 scoped files).

• Build ARSAS Windows x64 #38027414533: SUCCESS at exact head. Includes pinned-engine resolution, Release solution compilation, application regression tests, portable single-EXE publish, portable smoke test and artifact upload.
• Portable Windows x64 build #11660652579 — candidate for direct operator testing of the same-PC KM-TEST adapter case.
• Merge Execution Guard, Golden Provenance, Golden Budget Lock, Repeat-Run Stability, Production Promotion Guard, SV evidence: SUCCESS.
• Mainline Readiness #38027414490: expected failure / protected owing to outstanding R10 physical field authority; never disable guard.
• The separate field-capture packaging workflow may still be running; the successful Build ARSAS artifact above is already available.

Still required: test Windows NIC identity/capture on the operator's actual same-PC C264 simulator and confirm real 0x88B8 publisher frames (or identify a genuinely absent publisher). Identity-selected Npcap adapter is only observation readiness, not network evidence. The typed structured-FCD/report issue remains tracked in ARIEC61850 #159. Keep this PR DRAFT and do not merge/release. No manual CI reruns were performed.

[masarray]

BCUGE field result (2026-10-10): the user-tested merged candidate 95dcb82 (from #504+#505, engine #157) successfully captured 180/180 GOOSE frames, 0 other, across 3 live streams on same-PC KM-TEST with a decoded CSWI process change. Therefore P3A NIC selection is now LAB-VERIFIED in this configuration, not a generic network guarantee. Remaining defects are presentation/event coalescing, exact stVal/q reading, inspector ergonomics and NTP settings; draft #506 implements these separately. No report/control changes or release/mainline gate waiver.

[masarray]

Disposition: Superseded — verified stacked milestone consolidation; no mainline merge and no physical acceptance.

Canonical replacement: #543, open staged tip at 6285867780b8ea79c231d72a3819e9e5f99748d2. This PR HEAD 5e55fa2022bf3f79f1d747a176e75365306bdcd9 is a commit ancestor (verified GitHub compare: ahead=232, behind=0). Later changes in the stack can intentionally revise UI behavior; keep the canonical tip OPEN.

Unique delta: same-host smart NIC selection, capture scoping and GOOSE operator setup. Audited PR patch: 6-file patch contains GooseAdapterSelectionPolicy.cs, GOOSE_SUBSCRIBER.md and GooseAdapterSelectionPolicyTests.cs. Original changes and tests remain in reachable commit history of #543; no commit from this PR is orphaned. Historical ancestry alone does not certify current semantics, which remain subject to final exact-file and field validation at integration.

Safety: Only close the obsolete intermediate PR tracking record. The PR source branch, original commits, test/evidence history, engine lock, main, stable v1.6.42 and CI/physical gates are untouched. No release/engine authority claimed.

Follow-up: Keep NIC identity/source-binding evidence and same-host capture tests; no implicit change to physical subscription semantics. Canonical integration/qualification remains explicitly unfinished in #543 and successor release-gated workstreams.