[masarray]

Scope / provenance

• Based on unchanged main SHA 561bf69a5cceee76d66b46350decea1e7da7593d, published v1.4.1.
• Isolated v1.4.2 hotfix candidate; v1.4.1 tag and release artifacts remain untouched.
• Operator screenshots/handoff: single-click GOOSE publisher selection releases under snapshots; DataSet Inspector flickers/empties, timeline/publisher mix historical/live state.

Code-level root cause addressed

• Snapshot producer returns new cloned GooseMessageItem objects each snapshot; WPF ListView was bound to fully replaced IReadOnlyList and SelectedItem to a short-lived object. View recreation/null feedback could clear selection.
• Timeline selected-item setter also wrote publisher selection and unconditional collection-view Refresh() reset the DataGrid.
• Inspector network metadata came from a timeline event while DataSet values came from a different live publisher frame.
• Raw analyzer originally indexed publishers by GoCbRef alone; multiple physical publishers with the same control-block reference could coalesce.

Implementation

• Stable GoosePublisherKey: source MAC + APPID + VLAN + exact GoCBRef (fallback GoID/DataSet only if no GoCBRef); excludes stNum/sqNum/last-seen.
• Bounded keyed incremental ObservableCollection<GoosePublisherRow> with persistent row instances and change-specific property notifications.
• Selected publisher key is the single source for its latest frame; WPF transient null is not an operator Clear.
• Publisher click chooses Live latest coherent snapshot. Timeline click explicitly pins historical event. Timeline events are not allowed to mutate publisher selection; pure collection additions do not force view resets.
• STALE state holds selected publisher's last decoded typed DataSet; explicit Clear removes selection. Filter options only change when available identities change.
• Engine's raw GOOSE state key now separates same GoCBRef from distinct source MACs.
• Regression: 1000 retransmissions without row recreation; same GoID multi-source/VLAN; alternating traffic; temporary disappearance/recovery; state-number changes; explicit Clear; bounded 500-source churn; raw same-GoCb distinct-MAC test.
• No changes to 8-DIP layout, SV phasor, waveforms or PTP path.

Verification matrix

Automated: Windows CI, Runtime Stability, Runtime Architecture, CodeQL, candidate Windows/Inno install + repair, repository health at exact PR HEAD SHA. Runs are required before approval; a green CI does not constitute mouse-click proof.

Manual Windows acceptance still required: sustained real GOOSE stream, single click at 100%, 125%, 150% DPI; retransmissions OFF/ON; multiple publishers; changes of filter/search; select timeline then publisher; STALE/recovery; matching/missing SCL. Verify About BuildInfo exact commit and installed executable hash against candidate manifest.

CI efficiency

One focused PR branch (no redundant branch builds), version reserved pre-PR so the Inno package gate won't fail on the already-published v1.4.1. Normal concurrency cancellation handles superseded SHA builds; rerun only failed jobs after code changes. No release tagging or auto-merge.

Do not merge/release as validated until workflows are green and field selection acceptance is recorded.

[masarray]

Exact-SHA validation — v1.4.2 GOOSE selection hotfix candidate

PR HEAD: 44ec3fc0e041cada0125ede07c8b7985d5434bee (source branch fix/goose-publisher-selection-ssot-20261021; base main 561bf69a5cceee76d66b46350decea1e7da7593d).

All six checks completed SUCCESS for this exact SHA:

• Windows CI: repository health, WPF 8-DIP geometry, build, 193 passed / 0 failed / 0 skipped.
• Runtime architecture: PASS.
• Runtime stability: PASS.
• CodeQL: PASS.
• Dependency security: PASS.
• Windows Inno candidate: PASS, including package verification, Inno install + repair smoke and build provenance.

Candidate artifact: ProcessBusInsight-v1.4.2-win-x64-inno-candidate-292, GitHub artifact ID 12304818556; archive contains ProcessBusInsight-v1.4.2-win-x64-setup.exe, candidate-manifest.json, SHA256SUMS.txt, candidate-tests.trx. Candidate manifest commit is exactly 44ec3fc0e041cada0125ede07c8b7985d5434bee.

Installer SHA256 (verified directly against archive bytes): 5917a6fb2b61a6f706a7efd089c27c852cd119e4ea25e06b8591ffa5bfc2baba.

Important remaining gate — real Windows UI/traffic: while automated tests cover keyed reconciliation, real WPF mouse-click retention under sustained GOOSE retransmissions, selected-row highlighting at 100/125/150% DPI, switch to/from historical timeline, physical NIC/SCL mapping, STALE/resume and same-name multi-publisher behavior still need operator bench acceptance on the exact candidate EXE. CI cannot prove this. Keep PR draft and avoid publishing a stable release until that evidence is supplied.

No v1.4.1 release/tag/assets changed, no auto-merge, and no manual rerun/duplicate CI request was triggered.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-21T13:05:48.124543Z	44ec3fc	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.