Mark channel messages sent without scope

[dborup-agent]

Relates to #480

What

Channel messages sent without a region scope now show a compact amber ⚠ no scope marker where a scoped message shows scope: #dk, plus a Scope row with the same marker in packet detail for unscoped GRP_TXT/GRP_DATA. Scoped messages are unchanged. No API or Go change, no per-message requests.

Plan (small, single milestone)

1. public/app.js: isUnscopedFlood(routeType) (the one decision) and renderNoScope(routeType) (the one markup), next to isPathHashSizeWarn.
2. public/channels.js: meta line calls renderNoScope when there is no scope name and the route is not transport.
3. public/packets.js: scopeDetailHtml(pkt) keeps the existing scoped display and adds the marker row for unscoped GRP_TXT/GRP_DATA.
4. public/style.css: marker reuses --path-hash-warn (amber --warning, as Warn on 1-byte path hash: channel 'Sent with' badge, packet detail and node badge (recommend 2/3-byte) #353/feat(ui): warn on 1-byte path hash in channels, packet detail and node badge (#353) #356); no new colour.

Design decisions

• Detection is by route type only. FLOOD (1) is unscoped. TRANSPORT_FLOOD (0) is scoped, with or without a known name, so scope_name = '' is never "no scope". DIRECT (2) and TRANSPORT_DIRECT (3) get no marker.
• API shapes checked against the local server first: channel messages carry routeType (int) and scope (null when unscoped); packet detail carries route_type, payload_type and scope_name.
• Tooltip and sr-only text state only the firmware behaviour and the recommendation; nothing attributes a low heard-count to the missing scope.
• Perf: constant-time string building per message/row; no new loops, fetches or caches.
• Customizer: reuses the existing pathHashWarn token. A switch to hide the marker is a later customizer milestone (AGENTS.md rule 8).

Tests

• test-issue-480-no-scope.js (new, in test-all.sh): helper table, markup, channels meta-line render, theming.
• test-packets.js: 7 new cases for the Scope row.
• test-issue-480-no-scope-e2e.js (new, shard 2): real fixture message and GRP_DATA packet, scoped/unknown/direct/advert absence, light and dark, desktop and 360x800, axe color-contrast, no horizontal overflow.

🤖 Generated with Claude Code

[dborup-agent]

Rapport — CS-pve-agent2 PR#482 #480 — head d637cb3

Status: Implemented, all CI jobs green on head d637cb37. Draft, not merged. Needs review.

Evidence tags: [T] test run (local or CI), [A] analysis or code reading, [K] known or caveat.

Requirements

#	Requirement	Test (red before → green after)	Mutant (killed by)
1	Channels meta line shows ⚠ no scope (amber --path-hash-warn, warning icon, tooltip and sr-only text) for unscoped messages; scoped messages unchanged	test-issue-480-no-scope.js (16 new): aborts before (no helper), 16/16 after [T]. E2E test-issue-480-no-scope-e2e.js against master's frontend: 9 passed, 16 failed; on the branch 25/25 [T]	marker for every FLOOD/name-based/always → unit red [T]; text dropped, tooltip changed → red [T]
2	Packet detail Scope row with the same marker for unscoped GRP_TXT/GRP_DATA; scoped display kept	test-packets.js (7 new): 7 ❌ before → 194/194 after [T]. E2E covers a stubbed GRP_TXT and the real fixture GRP_DATA packet [T]	row dropped → E2E 8 failed [T]; payload gate removed or GRP_DATA dropped → unit red [T]
3	Detection by route type only: FLOOD = unscoped; TRANSPORT_FLOOD (named, empty name, null name) scoped; DIRECT/TRANSPORT_DIRECT, missing and invalid values get nothing	helper table in the unit file; E2E has scoped, unknown-name, null-name transport, direct and advert cases [T]. API shapes checked first: channel messages carry routeType (int) and scope (null); packet detail carries route_type, payload_type, scope_name [T]	TRANSPORT_FLOOD marked → unit 3 red and E2E 4 red [T]; string form dropped → red [T]. The first E2E version did not kill this mutant, so I added a null-name TRANSPORT_FLOOD case [T]
4	One shared helper (DRY)	isUnscopedFlood and renderNoScope in app.js, loaded as shipped by the tests; both call sites use them; a test asserts no inline markup in channels.js [A][T]	n/a
5	Theming, WCAG AA in both themes	no hardcoded colour in the .no-scope rule (unit) [T]; E2E checks computed colour = --path-hash-warn = --warning ≠ --danger, and axe color-contrast on every marker in light/dark, desktop/360x800 [T]	hardcoded colour → unit red; muted colour override → E2E 16 red [T]
6	Browser check, both themes, desktop and 360x800	E2E 25/25; no horizontal overflow, marker on-screen. Screenshots checked by eye for channels (fixture) and packet detail [T]	n/a

Invariants

• cmd/server read-only; the diff touches no Go files [T]
• No new map[string]interface{} [T]
• No hardcoded colours; scripts/check-xss-sinks.sh --diff origin/master clean, exit 0 [T]. The tooltip is a constant passed through escapeHtml [A]
• Fork guards unchanged: no workflow file is in the diff [T]
• No per-message API calls; constant-time string building [A]

Local runs (fixture prepared like CI: freshen, inline seed, migrate, seeds 2073/199/245/122)

• node test-frontend-helpers.js: 718 passed, 0 failed [T]
• eslint@8 on public/*.js: 0 errors [T]
• sh test-all.sh: 253 passed, 1 failed. The failure was test-test-all.js, because the new suites were not registered. Registered in test-all.sh and scripts/e2e-shards.json; test-test-all.js 8/8 and the shard tests pass after [T]. I did not rerun the whole test-all.sh afterwards [K]
• cmd/server go test ./...: ok [T]. cmd/ingestor not run locally because the PR has no Go diff; CI Ingestor job passed [K]

CI per job

Job	Result
Go Build & Test, Go Server Tests ×3, Go Ingestor, Go Libraries & CLI, Verify Go Test Assignments	pass [T]
JavaScript Unit Tests & Lint, Script & Container Unit Checks, Validate Docker Image	pass [T]
Playwright E2E 1/3, 2/3, 3/3, Verify E2E Shards & Coverage, Build & Publish Docker Image	pass on d637cb37 [T]
Release Artifacts / Publish Badges & Summary / Deploy Staging	skipping (fork-guarded) [T]

The first head, c64b52e0, failed E2E 3/3. My new marker carries the shared path-hash-warn class and sits in the same packet-detail list that the #353 E2E scanned for "2-byte must not warn", so 5 variants failed. This was a real interaction, not the known-flaky #271. Commit d637cb37 scopes that check to the Hash Size element (red 20/25 → green 25/25 locally). The Verify and Docker jobs had failed only as a consequence. No reruns.

Leftovers

• The hide/show choice for the marker is not in the customizer; it reuses pathHashWarn. A later milestone per AGENTS.md rule 8 [K]
• On staging data about 65% of channel messages are unscoped, so the marker will appear on most of them. It is compact, but the operator should see it live [K]
• The fixture has no scoped TRANSPORT_FLOOD channel message; scoped cases use channels.js test hooks and stubbed packet responses [K]

[dborup-agent]

Review — CS-MacBook PR#482 — head d637cb3

Dom: APPROVE med nits

Evidence tags: [T] test I ran, [A] code reading/analysis, [K] known caveat or not verified. Head checked with git ls-remote before and after: unchanged. Tested on the merged tree (git merge-tree of the head onto origin/master 31dac2b), served from a scratch archive.

Findings

#	Severity	Finding
1	nit	test-packets.js does not kill the "TRANSPORT_FLOOD marked" mutant (194/194 still pass). It is killed by the helper unit test (2 red) and by the E2E (4 red), so the criterion is covered, but the packet-detail unit table has no TRANSPORT_FLOOD row with scope_name: null [T]
2	nit	test-issue-353-path-hash-warn-e2e.js failed once on the head (node-detail contrast NaN:1, "fg on bg (none)") while Go suites were running in parallel. Three clean reruns: 25/25 each. Node detail is not touched by this PR and the check passed on the master frontend, so I treat it as timing flake [T][K]
3	nit	The marker has no customizer switch; reuses pathHashWarn. The author notes this as a later milestone (AGENTS.md rule 8) [A]
4	info	scopeDetailHtml keeps the old inline style="color:var(--text-muted)" for "unknown scope". It is pre-existing, moved unchanged, and uses a CSS variable [A]

No blocking findings.

Answers to the review points

• Detection by route type [T]: isUnscopedFlood is true only for 1 / '1'. TRANSPORT_FLOOD (0, with name, empty name, null name) is not marked; DIRECT (2) and TRANSPORT_DIRECT (3) are not marked; null/undefined/'x'/4/-1/true/{} are not marked. Channels: msg.scope → "scope: …"; transport routes → "scope: unknown"; only else the marker. Mutant "TRANSPORT_FLOOD marked" fails: unit 2 red, E2E 4 red (21 pass / 4 fail).
• API shapes [T]: local server on the fixture. GET /api/channels/%23wardriving/messages returns routeType (int) and scope (null) per message. GET /api/packets/<hash> returns packet.route_type, payload_type, scope_name (null for the seeded FLOOD GRP_TXT/GRP_DATA). The code reads exactly these fields. Caveat [K]: the fixture has no scoped TRANSPORT_FLOOD channel message, so scoped cases come from the test hooks and stubbed responses.
• DRY and style [A][T]: one isUnscopedFlood and one renderNoScope in app.js, next to the Warn on 1-byte path hash: channel 'Sent with' badge, packet detail and node badge (recommend 2/3-byte) #353 helpers. Both call sites (channels.js, packets.js) use them; a unit test asserts no inline "no scope" markup in channels.js. Same path-hash-warn class and warning icon (not error red). The only new CSS is .no-scope { white-space: nowrap }; no colour literal. E2E checks computed colour equals --path-hash-warn and axe color-contrast passes in light/dark at desktop and 360x800 (25/25).
• Tooltip [A][T]: text states only the firmware behaviour (default flood scope; flood.max.unscoped / region denyf may stop it) plus the recommendation. A unit test rejects "heard/repeats/lower/fewer". Mutant adding "so fewer repeaters hear it" fails 2 unit tests.
• Scoped messages unchanged / no per-message calls [A][T]: scoped and unknown-scope output is asserted byte-for-byte by name (scope: #dk, scope: unknown, "unknown scope"). The helpers are constant-time string builders; no new fetches.
• Browser [T]: the author's E2E ran by me against a local server: 25/25 on the head; against the master frontend 9 pass / 16 fail (red before, green after). I did not look at screenshots by eye [K].

Acceptance criteria → tests

Criterion	Test	Red before?
Unscoped message shows marker + explanation	test-issue-480-no-scope.js (16), E2E channels	unit aborts on master (helper missing) [T]; E2E 16 fail on master frontend [T]
Packet detail Scope row (GRP_TXT/GRP_DATA)	test-packets.js (7 new) + E2E	7 red on master public/ [T]
Route-type detection incl. empty name	helper table, channels render tests, E2E	mutants below
Scoped unchanged	unit + test-packets.js + E2E	mutant M3/M5
Contrast AA both themes	E2E axe	colour mutant below

Tests I ran (merged tree)

• cmd/server go test ./...: ok (61.9s) [T]
• cmd/ingestor go test ./...: ok (111s) [T]
• sh test-all.sh: 256 passed, 0 failed [T]
• node test-frontend-helpers.js: 719 passed, 0 failed [T]
• test-issue-480-no-scope.js 16/16, test-packets.js 194/194 [T]
• E2E on fixture (freshen, seed SQL, corescope-migrate, seeds 2073 and 199): test-issue-480-no-scope-e2e.js 25/25; test-issue-353-path-hash-warn-e2e.js 25/25 on reruns (see finding 2) [T]
• scripts/check-xss-sinks.sh --diff origin/master: exit 0 [T]
• CI on the head, per job: all Go, JS/lint, script, Docker, three Playwright shards and the shard/Go-assignment verifiers pass; Badges/Release/Deploy skipped [T]

My mutants

Mutant	Result
M1 isUnscopedFlood also true for 0 (TRANSPORT_FLOOD)	unit: 2 red; E2E: 4 red
M2 also true for 2 (DIRECT)	unit 3 red, test-packets.js 1 red
M3 scopeDetailHtml uses truthy scope_name (empty name falls to marker path)	test-packets.js 1 red
M4 tooltip appended with "so fewer repeaters hear it"	unit 2 red
M5 channels transport guard removed	unit 1 red
M6 E2E-served CSS overrides marker colour with muted	E2E 20 of 25 red

Invariants

• No Go files in the diff; cmd/server stays read-only; no new map[string]interface{} outside tests [T][A]
• No workflow file in the diff, so the fork guards are untouched [T]; I did not independently recount the 9 and 1 [K]
• No closing keywords (Relates to #480); commits are authored by dborup <kontakt@meshview.dk> [T]
• No hardcoded colours added [T]

Not verified

• Visual check by eye of screenshots (relied on E2E geometry, overflow and axe checks).
• Behaviour on real staging data (the author notes ~65% of messages will show the marker).
• Customizer exposure (out of scope for this PR).