Repository independence, provenance and maintainability hardening

[masarray]

Goal

Make the active ARSAS tree independently identifiable, reproducible and maintainable, without modifying the physical v1.6.40 release, claiming unverified copyright clearance, or falsifying third-party comparison provenance. This is the cross-cutting tracked maintenance program, not a request to rewrite Smart Discovery.

Completed or active scoped changes

• Exact triggering source checkout: fix(ci): verify exact triggering commit in Windows build and package #376 (merged).
• Retire historical stable-mutating publishers and make SBOM backfill additive: fix(release): retire obsolete 1.6.38 publishers and block SBOM overwrite #377 (merged).
• Neutralize two ARSAS-authored synthetic regression fixture/class names: refactor(tests): neutralize synthetic fixture and parity identifiers #378 (merged).
• Independent implementation and provenance contribution policy: docs(governance): independent implementation and auditable provenance #379 (CI pending as of creation).
• SCL semantic correctness candidate fix(scl): gate export-only counter correction on exact live integer type #374 remains a separate physical acceptance workstream under Post-release: finish SCL semantic parity without regressing accepted v1.6.40 #341.

Open acceptance checklist

• Inventory every Git-tracked path and textual content (source, test, workflows, scripts, evidence, docs, website/templates, legal notices, images/assets); do not rely on code search result counts as proof of absence.
• Classify every external product/brand identifier: unnecessary product-facing wording, internal name, synthetic fixture, honest black-box reference provenance, mandatory legal attribution, or potentially restricted asset. Record disposition per file, with a dated, reviewable manifest.
• Migrate the historical convergence workflow/doc/evidence to neutral active contract names atomically with all references, JSON keys, tests, event paths and check-name compatibility. Preserve exact request counts, the 58-member ordered authority, engine/app SHA, physical/test dates, historical failures and original source link via immutable commit. Do not claim measurements came from a different reference.
• Replace whole-file source-clean exclusions with narrowly justified, tested exceptions. Negative fixtures cover source, path, workflow, JSON, docs, extensionless policy files, split-token and mixed-case rejection; fix(source-clean): scan every tracked file and test rejection fixtures #384/perf(source-clean): execute full-tree fingerprint checks in compiled matcher #397/docs(provenance): make tracked asset inventory reviewable and CI-enforced #404 keep the scan exhaustive and fast.
• Audit field-compatibility names (including SNTP profile) separately; neutral SNTP commissioning profile symbols landed with packet bytes unchanged (f10cfcd).
• Audit third-party notices, licenses, bundled fonts/assets and website imagery; retain required attribution and independently produced fixtures. Review license consistency and sources, not keyword presence alone.
• Resolve historical PR fix(release): lock qualified runtime lineage and public surfaces #331, fix(discovery): preserve canonical static DataSet authority after live model build #333, discovery: converge live IED workflow with opened SCL static reporting #339, model: prepare source-neutral DataSet capability index and parity fingerprints #340 and perf(scl): reuse short-lived save enrichment snapshot #327 individually against current main, documenting unique changes before closure. Disposition audit is recorded in this issue; none is an authority to reintroduce stale Discovery/runtime or engine locks.
• Evaluate the separately scoped relaunch branch for private draft vs public site changes. Website changes are already identical to main; only the internal LinkedIn draft remains unique and is intentionally not restored. Deleting the branch would not erase Git/PR history.
• Retire remaining unused release/CI scripts only after verifying exact consumers and archived provenance; no workflow should mutate an existing stable tag.
• Protect main and required checks after verifying bot-driven website/evidence writes and external access controls; use a staged rollout.
• Refactor pure semantic/validation components under tests before association/monitoring or UI orchestration changes. Each PR states behavior-preserving invariants and measured effects.

Non-regression & completion

• v1.6.40 public tag/source/engine and binary SHA remain unchanged.
• No modification of Smart Discovery, MMS, static RCB/DataSet authority, cyclic poll quarantine or accepted 58/58 runtime through this hygiene program.
• Source clean, full Windows build/regression, deterministic guards, portable smoke, and post-merge production verification pass on the exact combined commit.
• Physical evidence remains attributable and reviewable, including rejected prior candidates.
• No "plagiarism-free", formal compliance or certified originality claim without an appropriate independent evidence review.

[masarray]

Maintenance integration update (2026-09-25): #384, #393, #394, #395, #396, #397, and #398 are now merged. The active tree has full tracked-path/text source-clean enforcement with negative fixtures (no whole-file exception), independent-implementation policy links and consistent public product notices, losslessly archived historical engineering audits, scoped Smart Discovery PR gates, grouped Dependabot updates, isolated WPF donor lifecycle smoke, and corrected documentation distinguishing Static DataSet Report Only from hybrid fallback. CI on #397 verified source-clean reduced from ~14 minutes to ~4 seconds without relaxing the checks; 1,312 app regression tests passed. The published v1.6.40 tag/source/assets and pinned engine remain unchanged.

Do NOT close this issue yet. Remaining work includes a reviewable per-file provenance/asset inventory, branch/history cleanup where authorized, staged main protection/required-check configuration (main is still unprotected), assessment of grouped dependency PRs #399/#400, and incremental behavior-preserving application refactors. Keep #374 under separate physical SCL acceptance in #341. No claim of certified originality or legal clearance is made by this maintenance work.

[masarray]

Maintainer progress 2026-09-25 (continued): #400 merged after 9/9 exact combined-head checks (grouped test-tooling dependencies); #401 merged after 10/10 checks, full 1,312/1,312 tests and build warnings reduced from 38 to 28 without production-code changes; #402 merged a snapshot-based tracked asset provenance register with 67 visual/font paths, 23 exact Git-blob duplicate groups, clear manual-review status for two legacy output images, and consistent ARSAS wording in the active clean-room policy. #403 is open for a bounded nullable diagnostic contract plus two tests, undergoing exact combined-head Windows CI. #399 remains open with explicit review note: grouped major GitHub Actions upgrades touch release/publishing paths not exercised by ordinary PR checks. Release v1.6.40 tag/source/assets and engine lock remain immutable.

Still open: human origin/rights and visual review of binary assets, staged main ruleset/branch protection, authorized orphan branch removal, targeted refactoring of pure logic under regression tests, and separate physical acceptance for SCL #374. A hash scan cannot certify originality or legal clearance.

[masarray]

2026-09-25 follow-up: #403 merged at 61ad333c2f61784ee49fda04b47f036206265cbc after 9/9 exact combined-head checks; it only corrects the nullable association-summary type contract and adds two deterministic diagnostic tests. Full Windows build reports 18 warnings, zero errors and 1,314/1,314 tests passed (versus 38 warnings and 1,312 tests before this maintenance pass). #400, #401 and #402 are also merged; #399 remains held for review of major GitHub Actions changes touching release/installer/Pages workflows, and #374 still requires independent physical SCL acceptance. Asset provenance register expressly leaves binary origin/rights and visual review open; no originality certification is claimed. Release v1.6.40 is unchanged.

[masarray]

Historical PR disposition audit against current main 514c151 (2026-09-25):

• perf(scl): reuse short-lived save enrichment snapshot #327 perf(scl): reuse short-lived save enrichment snapshot — closed/unmerged; 1 unique commit, now 159 commits behind. Its two-file optimization changes canonical-save reuse timing and must not be revived without a new exact-main benchmark/field qualification. No hygiene work depends on it.
• fix(release): lock qualified runtime lineage and public surfaces #331 fix(release): lock qualified runtime lineage and public surfaces — closed/unmerged; large historical integration branch (80 commits ahead, 139 behind) spanning runtime, engine pin, release/workflows, docs/site and tests. Later main independently contains the release-source hardening/public-surface cleanup. Do not merge or cherry-pick this stale aggregate branch.
• fix(discovery): preserve canonical static DataSet authority after live model build #333 fix(discovery): preserve canonical static DataSet authority after live model build — closed/unmerged; 18 commits ahead, 139 behind, including Discovery/runtime/static-reporting and an old engine lock. Its domain overlaps the subsequently field-qualified v1.6.40 lineage; do not revive it as authority or merge its stale engine/runtime changes.
• discovery: converge live IED workflow with opened SCL static reporting #339 discovery: converge live IED workflow with opened SCL static reporting — closed/unmerged; remaining unique diff is the old P0.7 release-candidate lock/evidence/test. It is historical candidate evidence, not a newer runtime baseline.
• model: prepare source-neutral DataSet capability index and parity fingerprints #340 model: prepare source-neutral DataSet capability index and parity fingerprints — closed/unmerged; 15 commits ahead, 119 behind, touching runtime/UI/model/value-formatting and tests. It remains an unqualified architectural experiment; any useful capability-index idea should be reimplemented as a fresh, bounded PR on current main rather than merged from this branch.

This audit intentionally does not claim Git history deletion and does not alter the accepted v1.6.40 source/tag/engine. #374 remains isolated under #341 for physical SCL acceptance.

[masarray]

Relaunch branch disposition (2026-09-25): docs/linkedin-relaunch-claims-2026-09-23 is 3 commits ahead / 27 behind current main. Its two website paths (landing/templates/id.html and landing/templates/index.html) are byte-for-byte identical to current main, so there is no unique website change to recover. The only unique tracked path is docs/LINKEDIN_RELAUNCH_2026-09-23.md, an editorial LinkedIn publication draft. Main already removed internal LinkedIn draft material from the public repository; do not merge/revive that document. Branch deletion would only reduce branch-list clutter and would not erase commit/PR history.

[masarray]

Release/CI consumer audit (2026-09-25): the primary packaging helpers build-windows-installer.ps1, create-windows-release-checksums.ps1, generate-release-sbom.py, publish-windows-portable.ps1, and sync-landing-release.py all still have active tracked workflow consumers. No script was deleted merely to reduce file count. PR #407 documents workflow ownership/mutation boundaries and removes stale hard-coded version examples. This checklist item remains open until the broader CI-script set is audited; current result is explicitly 'no safe deletion among the reviewed release helpers'.

[masarray]

Inter font provenance review (2026-09-25): official upstream release is https://github.com/rsms/inter/releases/tag/v4.1 and the recorded release archive SHA-256 9883fdd4a49d4fb66bd8177ba6625ef9a64aa45899767dde3d36aa425756b11e is corroborated by independent package records using the same official Inter-4.1.zip. The bundled face SHA-256 values are also corroborated for Regular 40d692fc...0c82, Medium 97ad806f...a872, SemiBold 78a843fa...cde3, and Bold 28831609...947f. PR #408 records this as technical byte provenance while retaining OFL-1.1 and explicitly not claiming legal clearance/originality.

[masarray]

Maintenance implementation started with draft PR #413 (head dc07b3f06fde71eb8f8370681f920850edd2eb8a, based on 9d081f54fc6466d13c099e8a94a020c487a36703). It addresses five nullable UI-facade device arguments, the stale source-version claim/two broken links in the docs index, and links contributor guidance to the already-existing AGENTS.md. Exactly three files; protocol, engine pin and release path unchanged. Exact-head Windows CI is running; no merge yet.

Next isolated CI-maintainability target (do not fold into #413): eight workflows trigger on the same small C# change. In at least smart-discovery-capture-build.yml, smart-discovery-golden-budget-lock.yml, smart-discovery-golden-provenance.yml, and smart-discovery-repeat-run-stability.yml, the checkout runs git clone --depth 1 --branch $ref using the mutable PR head branch name. Map each workflow's required head-vs-merge-candidate semantics, then use immutable expected SHA and prove exact checkout identity before consolidating any workflows. Keep the always-running Build ARSAS check and published v1.6.40 untouched.

[masarray]

Follow-up PR #414 is now open as draft at 06b0ed315893cb13b61ae8346fa383ccfe22fa37, based on current main 9d081f54fc6466d13c099e8a94a020c487a36703.

The four affected Smart Discovery workflows now use checkout of the exact triggering candidate followed by explicit git rev-parse HEAD == GITHUB_SHA verification, rather than cloning the mutable PR head branch name. Existing ArIED61850Tester paths, engine locks, fixtures and artifact names are unchanged. The other guards already using actions/checkout and the primary Build check were left untouched. Full CI is in progress; the Golden Budget Lock and Golden Provenance runs have already passed the new SHA-verification step. No main/release mutation, and no merge performed.

[masarray]

Maintenance continuation, 2026-09-26:

• PR chore(maintenance): clarify runtime device contract and documentation index #413 remains isolated and passed 8/8 exact-head workflows; it fixes five UI facade nullable call-site warnings and the documentation index.
• PR ci(maintenance): verify exact source SHA in four discovery guards #414 now passed 8/8 exact-head workflows. Four historical Smart Discovery jobs check out the triggering merge candidate and enforce git rev-parse HEAD == GITHUB_SHA instead of using mutable GITHUB_HEAD_REF.
• New draft PR ci(maintenance): use pinned ARIEC61850 in RCB export guard #415 (2dd0c8d20b88ee6a5f6b43688c646c24a117b6d9): Legacy SAS RCB export guard now uses the pinned ARIEC61850 SHA from engines/ARIEC61850.lock.json instead of upstream moving main, with lock-file path trigger and fail-closed checkout verification. Guard CI completed successfully and logged exact engine 648124097621046f5f127ceb1cf853fea54db730; full PR CI is still in progress.
• No change to published v1.6.40 or main, no auto-merge, no newly added vendor-specific runtime paths. DCO sign-off and exact combined-head validation remain required before integration.

Follow-on CI simplification should audit historical P0-5e/P0-5f workflows and required-check consumers before deleting or narrowing them; do not unilaterally disable project gates or assume green tests on independent PR heads substitute for the combined branch.

[masarray]

Combined-head maintenance validation is now represented by draft PR #416: c37bec2c20f99077d8b6b4b2d38ebcc55896e2c9 based on main 9d081f54fc6466d13c099e8a94a020c487a36703. It reproduces the eight exact source blob SHAs from #413 (3 files), #414 (4 files), and #415 (1 file) without additional edits/conflicts. The three constituent PRs are individually green (8/8, 8/8, 9/9); nine checks are triggered for the combined head. This validation-only PR remains draft, does not modify main/published v1.6.40, and must not be treated as DCO approval or auto-merged.

[masarray]

Integration follow-up: draft #416 is the consolidated first-pass maintenance candidate at c37bec2c20f99077d8b6b4b2d38ebcc55896e2c9, built from the exact eight changed file blobs of #413/#414/#415. The combined candidate passed 9/9 CI workflows, including Build ARSAS, Field Capture Build, the locked-engine selected-RCB guard and 1,315/1,315 application tests. Its PR body contains run links and remaining two trusted-SCL nullable-warning locations. It remains draft/unmerged because valid contributor DCO/CLA authorization and maintainer review are outstanding. Original source PRs have been cross-linked and remain open, to be closed as superseded only after integration. main and v1.6.40 were not changed.

[masarray]

2026-09-26 integration gate and next-subsystem progress:

1. chore(maintenance): consolidate runtime boundary and deterministic CI #416: exact 8-file consolidation still draft at c37bec2c20f99077d8b6b4b2d38ebcc55896e2c9; 9/9 exact-head workflows successful, 1,315/1,315 tests successful. main remains 9d081f54fc6466d13c099e8a94a020c487a36703. No PR reviews or DCO Signed-off-by trailer on the integration commit. CONTRIBUTING.md additionally requires affirmative agreement to CONTRIBUTOR-LICENSE-AGREEMENT.md. No affirmative DCO/CLA certification from the contributor has been verified. Do not merge or fabricate sign-off.
2. refactor(formatting): avoid second legacy scalar extraction #417: dependent, 2-file draft PR against chore(maintenance): consolidate runtime boundary and deterministic CI #416's integration branch; refactors Iec61850ValueFormatter so FormatReportProcessValue does not re-enter Format and attempt legacy structure parsing twice. Private helper retains the original scalar formatting switch. Added Boolean/Dbpos/indexed numeric/non-numeric structure regressions; source and tests only, no protocol or acquisition changes. CI triggered, pending. Do not merge until chore(maintenance): consolidate runtime boundary and deterministic CI #416 is legitimately integrated and refactor(formatting): avoid second legacy scalar extraction #417 is rebased/reconstructed on the actual final main with exact-head tests and contribution requirements satisfied.
3. The requested order remains #416 contributor certification → merge → verify main → #417 refactor integration; refactor(formatting): avoid second legacy scalar extraction #417 is preparation, not a claim that post-merge verification has happened.

No auto-merge, no changes to accepted v1.6.40 tag/artifacts or the pinned engine.

[masarray]

2026-09-26 formatter-subsystem validation completion: dependent draft #417 at 4fefd6bbf21f9e152c020dacebd134b76751c190 passed 9/9 exact-head GitHub Actions workflows. Windows Build ARSAS reports 1,319 regression tests passed, zero failed, zero skipped, versus 1,315 tests in #416. The two-file diff removes the duplicate legacy-structure extraction from Iec61850ValueFormatter.FormatReportProcessValue via a private shared normalized formatter, and adds four targeted scenarios. Audited real consumer boundaries: report-driven monitor and FAT preview vs generic direct reads/polling; TryNormalizeDbpos control/model consumers are unchanged. Two existing trusted-SCL nullable warning sites remain outside this change.

#416 is still a 9/9 green but draft/unmerged integration candidate; its commit has no DCO Signed-off-by trailer, there is no submitted review and no affirmative CLA agreement verified. The authorized contributor must satisfy DCO/CLA and review first. After #416 is legitimately merged, verify post-merge exact SHA, then restack/retest #417 against the real main merge commit. Do not treat these stacked PR results as post-merge verification. Published v1.6.40 source/engine/artifacts and main remain unchanged.

[masarray]

DCO handoff follow-up verified on 2026-09-26:

• chore(maintenance): consolidate runtime boundary and deterministic CI #416 was amended personally by the contributor to f7d1f65163eca3731c33056bcbe6aa8cbff5572b, carrying Signed-off-by: ARI SULISTIONO <ari.sulistiono@gmail.com>. Its tree is unchanged from c37bec2c20f99077d8b6b4b2d38ebcc55896e2c9: same eight modified files. Signed-off-head CI 9/9 success, including 1,315/1,315 tests and portable build.
• As of the direct GitHub issues/reviews checks, there is still no affirmative CLA agreement comment on chore(maintenance): consolidate runtime boundary and deterministic CI #416 and no submitted review. DCO is no longer the blocker for chore(maintenance): consolidate runtime boundary and deterministic CI #416. Do not fabricate or auto-post the contributor's legal acceptance.
• The SHA rewrite temporarily expanded stacked refactor(formatting): avoid second legacy scalar extraction #417's diff to 10 files. Recreated its exact original two-file tree on the new chore(maintenance): consolidate runtime boundary and deterministic CI #416 head without source-code changes: new b85ce12b4b75b5401be701ec4498b9fb261c0f99, exactly 2-file diff. New exact-head CI 9/9 success, 1,319/1,319 tests (4 added cases). Its own DCO sign-off/CLA gate remains for future merge.
• No PR merged, main/release v1.6.40 unchanged. Once the contributor personally records affirmative CLA acceptance for chore(maintenance): consolidate runtime boundary and deterministic CI #416 and maintainer review/integration requirements are met: merge chore(maintenance): consolidate runtime boundary and deterministic CI #416, verify actual main SHA and CI, then restack and requalify refactor(formatting): avoid second legacy scalar extraction #417 on that merged SHA.