Red commit: d761516d60 (no CI run —
branch-only push does not trigger workflows on this repo; verified
locally: `node test-issue-1668-m2-contrast.js` fails on assertion at
HEAD~1, passes at HEAD)
Partial fix for #1668 (M2 of 6). Fixes#1671.
## What changed
**Two-tier CSS tokens** introduced in `public/style.css`:
1. **Tier-1 (`--palette-*`)** — 38 raw colour stops, theme-independent,
in a single
`:root` block at the top of the file. Source: Tailwind v3 default
palette (MIT,
battle-tested for WCAG-graded luminance steps).
- gray ×9, blue ×9, green ×5, amber ×5, red ×5, purple ×5
- Single source of truth: no rule outside this block uses raw
`#hex`/`rgb()`.
2. **Tier-2 (semantic)** — existing `--text`, `--text-muted`,
`--surface-*`, etc.
re-plumbed to point at palette stops in both theme blocks. Behaviour
preserved
where contrast was already AA.
**WCAG AA bumps for M1 BLOCKER tokens**:
| Token / surface | Before | After | Theme |
|---|---:|---:|---|
| `--text-on-accent` on `--accent-strong` (was `#fff` on `--accent`) |
2.75:1 | **4.95:1** | dark + light |
| `--text-muted` on `--surface-1` | ~3.5:1 | **11.58:1** | dark |
| `--text-muted` on `--card-bg` | ~5.0:1 | **10.28:1** | dark |
| `--text-muted` on `#ffffff` | 5.74:1 | **10.31:1** | light |
| `--text-muted` on `--surface-0` | 5.32:1 | **9.45:1** | light |
**New tokens**: `--accent-strong` (= `--palette-blue-600` = `#2563eb`),
`--text-on-accent` (= `--palette-gray-50` = `#f9fafb`), `--text-subtle`.
Rules migrated to the new accent pair (all were `#fff` on
`var(--accent)` =
2.75:1 in the M1 audit): `.skip-link`, `.tab-btn.active`,
`.filter-bar .btn.active`, `.filter-group .btn.active`,
`[data-theme="dark"] .filter-bar .btn.active`, `.path-hops .hop-named`.
## Operator-reported chip (2026-06-12 — `.hop-named.hop-link`)
Dark-blue text on dark-blue chip background. Patched via `.path-hops
.hop-named`
+ `--accent-strong`. Now reads as `#f9fafb` on `#2563eb` = 4.95:1 (AA
pass) in
both themes. Before/after screenshots: see
`a11y-audit/m2-screenshots/{before,after}-packets-{dark,light}-1200x900.jpg`.
Letsmesh's UI uses chip text at 6.77:1 on equivalent surfaces; this PR
closes
most of the gap.
## TDD trail
- Red commit `d761516d` — assertion-based contrast test, fails on
missing palette.
- Green commit `e5e87309` — palette + remap + bumps + AA pass.
- Anti-tautology: reverting dark `--text-muted` back to `#6b7280`
reproduces
`text-muted on surface (dark): contrast 3.53:1 < 4.5:1`.
## Preflight
`bash ~/.openclaw/skills/pr-preflight/scripts/run-all.sh origin/master`
→ all
gates pass (incl. CSS-var-defined: 1928 var() refs, 0 undefined).
## Not in this PR (intentional)
- M3 typography (`14px` floor + weight 500 for chips/badges) — own PR
- M4-M5 per-route polish — own PRs
- M6 axe CI gate — own PR
- Shipping an alternate palette — deferred, indirection enables it
---------
Co-authored-by: openclaw-bot <bot@openclaw.dev>
Release notes for v3.8.4 — the "Phosphor migration" release. Six PRs
(#1649–#1654, tracking #1648) plus three followup fixes
(#1659/#1660/#1665) replaced all decorative emoji in the UI with
Phosphor sprites and added a lint gate to prevent regression.
## Verification summary
Test plan: `workspace-meshcore/test-plans/v3.8.4-cdp-test-plan.md` (93
tests, 16 sections).
- Initial run (pre-#1665): 56 pass / 22 partial / 5 fail / 14 skipped.
Two BLOCKER lint-gate breaches in observers and analytics Channels.
- Final run (post-#1665, hot-patched to staging): both blockers ✅ —
v384-1.2 (11 chips, 11 sprites, 0 emoji), v384-12.18 (315 lock sprites,
0 🔒 emoji).
- 22 partials are plan selector drift, not code regressions; deferred to
v3.8.5.
## Tagging
Per the notes file, this is ready for `git tag -a v3.8.4 037dc8c4 -m
"v3.8.4"` after merge — **not executed by this PR**.
## Review
Draft for user review. Will be marked ready / merged before tag.
---------
Co-authored-by: meshcore-bot <bot@meshcore.dev>
## What
Two Phosphor lint-gate breaches found by the v3.8.4 manual-test executor
— app-controlled UI labels still shipping raw emoji glyphs that the M6
final sweep (#1648) missed. One PR, two sprite swaps, same playbook as
#1657.
### Findings
| Test ID | Surface | Glyph | File:line | Fix |
|---|---|---|---|---|
| v384-1.2 | `/observers` `.obs-clock-naive-chip` | `⚠️` (U+26A0) ×14 |
`public/observers.js:30` | `ph-warning` sprite |
| v384-12.18 | `/analytics?tab=channels` encrypted row name cells | `🔒`
(U+1F512) ×158 | `public/analytics.js:978–979` | `ph-lock` sprite |
Finding 2 is a different surface from the M3/#1657 fix (which swapped
the section-header label, not the per-row `displayName`). The
unknown-encrypted row's `displayName` carried a raw `🔒 Encrypted (0xNN)`
text label that then flowed through `esc()` into the rendered name cell
as an escaped emoji glyph — exactly the same `innerText → innerHTML`
class of bug. Refactored to mirror the section-header pattern:
`displayNameHtml` carries the sprite-bearing raw HTML; `displayName`
stays plain text for sort/aria/tests.
## TDD
- **RED** `cde12370` — `test-issue-1648-followup-phosphor-leaks.js`
asserts ph-warning sprite + zero ⚠ in chip output, and ph-lock sprite +
zero 🔒 in analytics row labels. 6 assertions failed on master.
- **GREEN** `f1c64b17` — sprite swaps applied. All 9 assertions pass.
- **Anti-tautology proven both directions**: reverting only
`public/observers.js` → 2 chip-related assertions fail; reverting only
`public/analytics.js` → 4 analytics-related assertions fail.
## Verify
- ✅ `node test-issue-1648-followup-phosphor-leaks.js` — 9/9 pass
- ✅ `node test-issue-1648-m6-final-sweep.js` — 0 violations
- ✅ `node test-observer-naive-clock-1478.js` — 8/8 pass (existing chip
test accepts ph-warning sprite)
- ✅ `node test-analytics-channels-integration.js` — pre-existing
unrelated `Channel Analytics` failure only; encrypted-row assertions all
pass with new plain-text `displayName`
- ✅ pr-preflight all gates green (PII, branch-scope, red-commit,
CSS-var, LIKE-on-JSON, async-migration, XSS sinks)
- ✅ Browser-verified on staging: 11 chips render ph-warning sprite (0
emoji), 156 ph-lock sprites in row name cells (0 lock emoji on page)
Browser verified: http://analyzer-stg.00id.net/#/observers +
/#/analytics?tab=channels (hot-patched)
E2E assertion added: `test-issue-1648-followup-phosphor-leaks.js:67`
(chip), `test-issue-1648-followup-phosphor-leaks.js:147` (row cell)
---------
Co-authored-by: meshcore-bot <bot@meshcore.dev>
## Summary
Fixes the ~5% flake in `test-slideover-1056-e2e.js` packets@800 subtest
caused by a virtual-scroll spacer race.
## Root cause (per issue)
The packets `PAGES` entry used a loose selector with a bare-`tr`
fallback (`#pktTable tbody tr[data-id], #pktTable tbody tr`). That
fallback matches the virtual-scroll spacer `<tr>` (no `data-*`, no click
handler). The row-wait guard counted any `<tr>`, so it was satisfied by
the spacer alone — the test then clicked the spacer and no slide-over
opened.
## Fix (test-only)
1. `test-slideover-1056-e2e.js` L42 — drop bare-`tr` fallback; use
`#pktTable tbody tr[data-id]` only.
2. `test-slideover-1056-e2e.js` L69-71 — `waitForFunction` now queries
the page-specific `rowSel` directly (`querySelector(rowSel) !== null`)
instead of counting any `<tr>`. Works for all three pages (packets /
nodes / observers) because each already uses an attribute-strict
`rowSel`.
## TDD
- Red commit `d02e496b`: new `test-slideover-1056-rowsel-strict.js` pins
the discipline — fails when the bare-`tr` fallback or loose `tbody tr`
count is present.
- Green commit `a8b445d5`: applies the selector + guard tightening; pin
test passes.
## Verification
- `node test-slideover-1056-rowsel-strict.js` → 2/2 pass on the fix
commit; 0/2 on parent.
- `node -c test-slideover-1056-e2e.js` → syntax OK.
- Preflight (`bash ~/.openclaw/skills/pr-preflight/scripts/run-all.sh
origin/master`) → all gates green.
Scope is strictly test-only — no production code touched.
Fixes#1662.
---------
Co-authored-by: clawbot <bot@corescope>
Co-authored-by: clawbot <bot@local>
Co-authored-by: clawbot <clawbot@users.noreply.github.com>
Master CI's Go test job has been timing out at the default 10 minutes
since #1655 (`825b2648`) landed additional endpoint-coverage + race
tests. This bumps the explicit `-timeout` on both `cmd/server` and
`cmd/ingestor` test steps to 15 minutes.
No code/test changes — config-only. This is preventative; the slow tests
are a separate follow-up.
### Timing data (local sandbox, arm64, slower than CI)
| Module | Duration |
|---|---|
| `cmd/server` (`go test -race ./...`) | **9m 30s** — already grazing
the 10m default |
| `cmd/ingestor` (`go test ./...`) | 2m 44s |
The server suite is now consistently above 9 minutes and any test added
on top of #1655 pushes it past 10m on slower CI runners (the failure
mode we hit on master).
### Change
```diff
- go test -race -coverprofile=server-coverage.out ./...
+ go test -timeout 15m -race -coverprofile=server-coverage.out ./...
- go test -coverprofile=ingestor-coverage.out ./...
+ go test -timeout 15m -coverprofile=ingestor-coverage.out ./...
```
Out of scope: optimizing the slow tests
(TestGetChannelMessagesPerfLargeChannel etc.) — separate issue/PR.
Co-authored-by: corescope-bot <bot@corescope>
## Problem
A relay node's **activity timeline** — and its per-node `packetsToday` /
observer counts — collapses to *"only the hour the server restarted"*
after every restart. Before the restart the timeline shows only the
node's own adverts (~1–2/hr); all of its relay activity piles into the
single post-restart hour.
## Root cause
All DB cold-load paths (`Load`, `loadChunk`, `scanAndMergeChunk`) index
relay-hop attribution into `byNode` **only** from
`observations.resolved_path`. But since #1287 the ingestor persists
relay data as aggregate `neighbor_edges` and **never writes
`resolved_path`** — it is `NULL` on every deployment (verified on a live
DB: 0 of ~440k rows populated). So relay attribution is never
reconstructed on startup; it only re-accumulates from live traffic
(`IngestNew*`, which re-resolves from `path_json` + the neighbor graph),
piling a relay node's whole history into the post-restart window.
## Fix
Server read-side only — **no schema / ingestor / migration change**.
When `resolved_path` is empty, re-resolve relay hops from the
already-persisted `path_json` using the in-memory prefix map + neighbor
graph (the same `resolvePathForObs` compute the live ingest path already
runs). `main.go` now loads the persisted neighbor graph *before* the
packet load so resolution has the graph available.
Two correctness details worth a close look:
1. **Fetch the prefix-map/graph snapshot BEFORE opening each load
cursor.** `getCachedNodesAndPM` issues its own DB query; doing so while
a load cursor is open deadlocks on a single-connection SQLite pool (the
test harness uses one).
2. **Index into `byNode` ONLY** — not the `resolved_path` / path-hop
indexes. Those are cross-checked by `handleNodePaths` against the
persisted `resolved_path` column (NULL here); populating them from an
in-memory re-resolution would make that SQL confirmation fail and
wrongly drop the tx from paths-through (#1352).
## Tests
New coverage asserts a relay pubkey reachable *only* via `path_json`
lands in `byNode` after a restart-style load, for both the hot-window
(`LoadChunked`) and background-window (`loadChunk`) paths. Existing
#1558 (`resolved_path`) and #1352 (paths-through) tests still pass. Full
`cd cmd/server && go test ./...` is green under `-race`.
## Perf
The fallback runs `resolvePathForObs` per observation with a non-empty
`path_json` during cold load — the same per-packet compute the live
ingest path already performs, so no new asymptotic cost. The prefix map
+ graph are snapshotted **once per load** (not per row);
`getCachedNodesAndPM` is 30s-cached. In `loadChunk` the resolution runs
in the existing lock-free scan and is accumulated locally, matching that
function's "build local, merge under lock" design.
## Note on a pre-existing flaky test
`TestDistanceConcurrentRequestsDuringBuildReturn202` is timing-fragile
(fails ~1/15 on `master` without this change). It relies on the lazy
distance build being slow because it's the first caller of
`getCachedNodesAndPM` (cold cache). This PR pre-warms that cache during
`Load`, narrowing the build window, so the test fails more often in
**non-race** local runs. It passes reliably under `-race` (CI mode),
where the build stays slow. Flagging in case you want to harden the test
separately.
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: openclaw-bot <openclaw-bot@users.noreply.github.com>
Co-authored-by: openclaw-bot <bot@openclaw>
Fixes#1181.
## Summary
Adds operator-configurable name-prefix hiding for nodes. When a node's
name starts with any prefix listed in the new `hiddenNamePrefixes`
config field (default `["🚫"]`), it is omitted from `/api/nodes`,
`/api/nodes/search`, and `/api/nodes/{pubkey}`. DB rows are preserved —
the filter runs at the API layer only, so observation history (paths,
hops, distances) stays intact and the node simply re-appears if the
operator clears the prefix list.
This mirrors the convention already in use on other MeshCore map
dashboards: an operator who wants their node hidden renames it with the
🚫 prefix and sends an advert; the next advert is then dropped from the
dashboard. The node is **not** hidden from the mesh itself — only from
this dashboard. This is documented inline in `config.example.json`.
Implementation follows the existing `IsBlacklisted` pattern exactly: a
new `Config.IsNameHidden(name)` method, and three filters in `routes.go`
placed alongside the corresponding blacklist filters. No DB schema,
public API, or websocket changes.
## Files changed
- `cmd/server/config.go` — new `HiddenNamePrefixes []string` field +
`IsNameHidden` method
- `cmd/server/routes.go` — filters in `handleNodes`, `handleNodeSearch`,
`handleNodeDetail`
- `config.example.json` — new field + `_comment_hiddenNamePrefixes`
operator doc
- `cmd/server/hidden_name_prefix_1181_test.go` — new test file (red →
green)
## Test plan
Two new subtests in `TestHiddenNamePrefix_1181_*`:
1. `_NodesList` — inserts a node named `🚫 ban me`, asserts it is present
when `HiddenNamePrefixes` is empty and absent when set to `["🚫"]`.
2. `_Search` — inserts `🚫 search me`, asserts
`/api/nodes/search?q=search` does not surface it when the prefix is
configured.
Verified red→green:
- Red commit `d0903852`: `go test -run TestHiddenNamePrefix_1181` fails
on the leak assertion (`hidden_name_prefix_1181_test.go:94`).
- Green commit `e79a0d8d`: same command passes.
```
$ cd cmd/server && go test -run TestHiddenNamePrefix_1181 -count=1 .
ok github.com/corescope/server 0.060s
```
## Out of scope
- Auto-purging DB rows for hidden nodes — left to existing retention.
The triage was explicit: hide, do not delete.
- Live websocket broadcast: nodes are not broadcast via websocket (only
packets), so no separate emit path needs filtering. Frontend reads nodes
via `/api/nodes`, which is filtered.
- Frontend customizer for the prefix list — operators configure via
`config.json` like every other knob.
Fixes#1323
## Summary
Adds a small in-memory cache of the community-maintained
hashtag-channels
catalogue (`marcelverdult/meshcore-channels`) and exposes it as
`GET /api/known-channels?region=XX` plus a collapsed sidebar section on
the Channels view ("Known channels (catalogue)") with a one-click
"+ Add" button per row.
Per triage (#1323): new `cmd/server/known_channels_cache.go`, new
`GET /api/known-channels?region=…`, frontend section in
`public/channels.js`. No new DB tables — cache is in-memory only.
## What changed
- `cmd/server/known_channels_cache.go` — `knownChannelsCache` with an
atomic snapshot pointer, 24h default refresh, 30s HTTP timeout, 4 MB
body cap, custom `User-Agent`. Fail-soft: a failed refresh leaves the
last-known snapshot in place. Background goroutine started from
`main.go` after the neighbor-graph recomputer; never blocks startup.
- `cmd/server/known_channels_route.go` — `GET
/api/known-channels?region=`
serves the cached snapshot off the atomic pointer (never blocks on
upstream). Region filter is case-insensitive ISO 3166-1 alpha-2.
Empty/missing cache returns 200 with an empty entries list (fail-soft
for the UI).
- `cmd/server/config.go` — `KnownChannelsURL` +
`KnownChannelsRefreshMs`.
- `config.example.json` — example values + `_comment_knownChannels`.
- `public/channels.js` — new collapsed sidebar section "Known channels
(catalogue)" that lazy-fetches `/api/known-channels` on first render
and renders rows with a "+ Add" button. The button calls the existing
`addUserChannel(name)` path, so adding catalogue channels reuses the
full save-key + decrypt flow that user-typed hashtags already use.
- `cmd/server/known_channels_cache_test.go` — failing-first tests:
- `TestKnownChannelsParseFixture` asserts the parser populates
`GeneratedAt`/`License` and region-stamps every entry while skipping
empty countries.
- `TestKnownChannelsRouteRegionFilter` asserts the route returns 200
with exactly the filtered subset for `?region=be`.
- `TestKnownChannelsFailSoftOn500` asserts a failed upstream fetch
leaves the prior snapshot in place and bumps `failCount`.
## Upstream pinning
The default URL is pinned to the specific file
`channels-by-country.json`
on `main`:
>
https://raw.githubusercontent.com/marcelverdult/meshcore-channels/main/channels-by-country.json
Shape (verified 2026-05-24):
```json
{
"generated_at": "...",
"license": "CC0-1.0",
"countries": { "be": [{"channel": "#antwerpen", "description": "..."}], ... }
}
```
## Test plan
```
cd cmd/server && go test -run 'TestKnownChannels' -count=1 .
ok github.com/corescope/server 0.008s
```
Red commit: 5c43cff3 (all three tests fail on assertions, build clean).
Green commit: 54a1080e (parser + cache + route implemented, all three
pass).
## TDD evidence (red → green)
- **Red commit `5c43cff3427afd8aa2f3cce20c31058190aebc37`** — tests
added
with stub implementations that compile but return zero/empty so each
test fails on an assertion (not a compile/import error). `go test -run
TestKnownChannels` output captured in the commit message.
- **Green commit `54a1080e45fd2e10da2caa156f376bf4d0212976`** — parser,
cache, route, main-wiring, frontend section land; all three tests
pass.
## Frontend verification
Browser verified: http://analyzer-stg.00id.net/#/channels (with the
`/api/known-channels` response stubbed in DevTools to simulate the cache
being populated on staging, which is still on master and doesn't have
the new endpoint yet).
E2E assertion added: cmd/server/known_channels_cache_test.go:71 —
asserts the route returns 200 and the response body's `entries` length
matches the filtered subset.
## Limitations / follow-ups (not in scope of this PR)
- The catalogue only ships PSK keys for a small subset of entries (the
upstream schema makes `key` optional). For entries WITHOUT a `key`,
the "+ Add" button still wires through `addUserChannel("#name")` —
which derives the standard public-channel key from the name (the same
path used today when a user types `#foo` into the Add Channel modal).
For entries WITH a `key`, a follow-up PR can pass the key through to
`addUserChannel` so the UX matches "paste-a-PSK". Today the key is
shown in the JSON payload but not yet wired into the FE button.
- No deduplication against the in-memory `/api/channels` list — the
catalogue section is intentionally separate so the user sees which
channels exist worldwide even if their server hasn't seen traffic.
- No per-section region selector yet — the section shows the full
catalogue regardless of the page-level region filter. Future work:
add a dropdown.
## Preflight
```
═══ Preflight clean. ═══
```
cross-stack: justified — issue #1323 spans `cmd/server` (cache + route)
and `public/channels.js` (sidebar surface); same feature, both halves
required.
---------
Co-authored-by: Kpa-clawbot <bot@corescope.local>
Fixes#1657
## Bug
On `/analytics` → **Channels** tab, the "Channel Activity" table's
group-header rows ("My Channels", "Network", "Encrypted") rendered
literal HTML source text:
```
<SVG CLASS="PH-ICON" ARIA-HIDDEN="TRUE"><USE HREF="/ICONS/PHOSPHOR-SPRITE.SVG#PH-KEY"/></SVG> My Channels
```
instead of the actual Phosphor sprites. Per-row encrypted/lock icons
rendered fine — the bug was isolated to the group-header render path.
## Root cause
`public/analytics.js` `channelTbodyHtml` builds each group-section
header by wrapping the section label in `esc()`:
```js
esc(sections[si].label) + ' <span class="text-muted">(' + rows.length + ')</span>'
```
But the labels (`sections[].label`) are hardcoded sprite-bearing
strings:
```js
{ key: 'mine', label: '<svg class="ph-icon" aria-hidden="true"><use href="…#ph-key"/></svg> My Channels' },
```
`esc()` HTML-encoded the `<` / `>` so the browser displayed the source
text rather than rendering the sprite. Affects all 3 groups (and any
future group with a sprite).
## Fix
Drop the `esc()` wrap on the hardcoded label (single line change, same
pattern as M3 commit 4ca73ced for mobile channel avatars). The
`(<count>)` suffix is numeric and was always safe.
## Tests
New `test-issue-1657-analytics-channels-group-sprites-e2e.js` (mobile
375 viewport, matching the bug report):
- (1) at least one group-header row renders
- (2) every header row contains a real `<svg.ph-icon>` child
- (3) per-group sprite refs resolve (My Channels → `#ph-key`, Network →
`#ph-radio`, Encrypted → `#ph-lock`)
- (4) the Channel Activity table's `innerText` contains no literal
`<svg` substring (escape-leak gate)
Wired into the CI E2E lane (`.github/workflows/deploy.yml`) immediately
after the M4 icons E2E.
## TDD evidence
- Red commit: `8f8781c1` — test + CI wiring only, no production change.
Test asserts behavior that did not exist on master → CI fails on this
commit.
- Green commit: `8385fa54` — 1-line fix to `public/analytics.js`. Test
passes.
## Anti-tautology proof
Hot-patched staging with the `analytics.js` from the red commit
(pre-fix), reloaded `/analytics?tab=channels` at 375 viewport, and the
in-browser DOM probe returned:
```
headers[0].text = "<svg class=\"ph-icon\" aria-hidden=\"true\"><use href=\"/icons/phosphor-sprite.svg#ph…"
headers[0].svgs = 0 // (2) would fail
headers[1].svgs = 0 // (2) would fail
literalSvg = true // (4) would fail
```
Restored the fixed file; same probe returned `svgs=1`, correct `uses[]`
refs, `literalSvg=false`.
## Staging verification
Hot-patched `corescope-staging-go:/app/public/analytics.js` (no restart
needed — static file). Mobile dark @ 375 viewport shows Network → radio
sprite and Encrypted → lock sprite rendering correctly. (My Channels
group not present because the e2e fixture has no `mine`-tagged channels
— expected; the test skips that assertion when the row is absent.)
## Scope discipline
Touched only:
- `public/analytics.js` (1-line `esc()` removal + comment)
- `test-issue-1657-…-e2e.js` (new)
- `.github/workflows/deploy.yml` (1-line E2E wire)
No broadening, no helper renames, no related-but-different
escape-removal opportunism.
---------
Co-authored-by: openclaw-bot <bot@openclaw.local>
Draft for milestone 4 of #1648 — emoji → Phosphor Icons (map & route
overlays).
Currently at the red commit (failing test only). Implementation follows.
Partial fix for #1648 (M4 of 6). Do NOT close the tracking issue.
---------
Co-authored-by: bot <bot@corescope>
## Summary
Polish follow-ups for the #1644/#1645 observer-comparison redesign —
addresses all 5 parent visual-review findings + 3 Tufte additions in one
PR. Fixes#1646.
## Coalesced fix list (status: ✅ all landed)
| # | Tag | Item | Fix | Evidence |
|---|---|---|---|---|
| 1 | [both] | Native checkboxes were bare white squares against dark
theme | Global `input[type=checkbox]/[type=radio] { accent-color:
var(--accent) }` + `color-scheme: dark` on dark theme blocks. UA renders
themed checkboxes everywhere now. |
`screenshots-1646/after-observers-selected-dark.jpg` vs
`before-observers-selected-dark.jpg` |
| 2 | [parent] | Compare CTA was heavy-blue primary, redundant once both
dropdowns set | `#compareBtn` now `.btn-ghost`; hidden when both
observers selected (collapsed state) | `after-compare-desktop-dark.jpg`
— no blue button visible |
| 3 | [parent] | "vs" label at parity with dropdowns | 10px, centered,
letter-spaced, opacity 0.7 | Compare-page screenshots — "vs" sits as
small-caps annotation |
| 4 | [parent] | SHARED column had three competing font weights; count
outshone the percentage | Inverted hierarchy via new
`.compare-strip-mid-pct` + `.compare-strip-mid-pct-unit`. 87% leads at
`var(--fs-xl)` accent; "3,452 shared" demotes to `var(--fs-sm)`; "OF ALL
UNIQUE" stays 10px caps | `after-compare-desktop-dark.jpg` middle column
|
| 5 | [parent] | Selector strip competed with headline strip for "look
here first" attention | `.compare-controls.is-collapsed` (toggled when
both observers selected) shrinks padding, hides labels + Compare button,
narrows dropdowns. A/B swap still reachable |
`after-compare-desktop-light.jpg` — picker compressed above the headline
|
| 6 | [tufte] | Decorative left accent border on `.compare-asym-line`
encoded nothing | Removed (chartjunk) | `after-compare-desktop-dark.jpg`
— squared cards |
| 7 | [tufte] | Decorative left green border on `.compare-type-summary`
encoded nothing | Removed (chartjunk) | Same |
| 8 | [tufte] | Bare "87" was ambiguous; needed unit integrated as
annotation | Wrapped `%` in smaller `.compare-strip-mid-pct-unit` —
words and graphics co-located | Middle column hierarchy |
## Push-backs / scope discipline
- **Did NOT remove the selector strip entirely.** Parent rule + operator
UX: A/B swap must remain reachable. Collapsing > removing.
- **Did NOT introduce a custom checkbox widget.** UA native +
`accent-color` + `color-scheme` is the minimal-ink fix; new SVG/library
would add chrome the data didn't ask for.
- **Did NOT add new color tokens.** All restyling uses existing
`--accent`, `--text`, `--text-muted`, `--surface-1/2`, `--border`.
## TDD
- Red commit: `2863cfb3 test(#1646): RED — assertions for compare-polish
...` — `test-issue-1646-compare-polish.js` 9 assertions, all FAIL on
master.
- Green commits (3 logical groups):
1. `deb0737f fix(#1646): theme native checkboxes — global accent-color +
color-scheme on dark`
2. `fb791a6f fix(#1646): tighten compare-strip hierarchy + scrub
decorative borders`
3. `8033ac36 fix(#1646): ghost-style Compare CTA + collapse the picker
once both observers chosen`
Final: `node test-issue-1646-compare-polish.js` → 9/9 pass; `node
test-issue-1644-redesign.js` → 13/13 pass (no regression); `node
test-compare-overlap.js` → 6/6 pass; `node test-frontend-helpers.js` →
611/611 pass.
## Visual verification
All staging-validated via local headless chromium against the
hot-swapped files at `http://20.x.y.z` (staging). Surface matrix
covered:
- Observers list — desktop dark (with 2 rows checked) — themed accent on
checkboxes
- Observers list — mobile 375px dark
- Compare page — desktop light + dark
- Compare page — mobile 375px dark
**Reviewer note: screenshot artifacts were captured locally (sandbox
does not have a GitHub UI session for attachment upload).** Paths below
— pull these from the same workspace location if you want to inspect:
```
screenshots-1646/before-observers-desktop-dark.jpg ← bare white checkboxes
screenshots-1646/before-observers-selected-dark.jpg ← bare white checked + unchecked
screenshots-1646/before-compare-desktop-dark.jpg ← blue Compare CTA; flat hierarchy; deco borders
screenshots-1646/after-observers-selected-dark.jpg ← themed checkboxes
screenshots-1646/after-observers-mobile-dark.jpg
screenshots-1646/after-compare-desktop-light.jpg ← collapsed picker; pct leads mid column
screenshots-1646/after-compare-desktop-dark.jpg
screenshots-1646/after-compare-mobile-dark.jpg
```
No raw `MEDIA:` UUIDs in this body — that was the mistake on #1645 and
is not being repeated. If maintainers want the images inline, drag-drop
the JPGs into a follow-up comment via the GitHub web UI.
## Risk
Low. Pure CSS + one class-toggle in `compare.js`'s `updateBtn`
(idempotent, no race, no event loop change). `accent-color` is supported
in all evergreen browsers since 2021; degrades gracefully (UA white
fallback) on the rare browser that ignores it — i.e. exactly the
current-master state.
---------
Co-authored-by: openclaw-bot <openclaw-bot@users.noreply.github.com>
Co-authored-by: openclaw-bot <bot@openclaw.local>
Co-authored-by: clawbot <clawbot@users.noreply.github.com>
## What was wrong
PR #1642 promoted observer comparison to a first-class IA citizen but
shipped three problems: `class="btn-secondary"` buttons that fell back
to browser-default white/gray because no such CSS rule existed; the
30-second auto-refresh blew away `<tbody>.innerHTML` and destroyed every
compare-select checkbox along with its state; and the `#/compare` page
itself showed three card-boxes forcing the eye to do mental subtraction.
## Design rationale (Tufte)
The comparison page now leads with **one row of three numbers above
one proportional diff bar** — shared-axis small multiples in place of
three nearly-identical cards. The eye reads the whole comparison in
one fixation. Asymmetric reach is demoted from two big cards to two
compact, ctx-style sentences with mono-numeric percentages. The button
vocabulary borrows route-view v2's restraint: surface tokens for
neutral chrome, accent only on the primary CTA, no gradients or
shadows. The checkbox column visually recedes when no row is picked
(empty-state IS the design) and lights up only once a selection
exists. Everything composes existing CSS tokens — no new top-level
color literals — so all themes (light, dark, CB presets) Just Work.
## Inventory of CSS additions
| Selector | Role |
|---|---|
| `.btn-secondary`, `.btn-secondary[disabled]` | Themed neutral button
(low-emphasis CTA) |
| `.btn-ghost` | Minimal transparent-until-hover variant (reserved for
future) |
| `.compare-page`, `.compare-page .page-header` | Page-level container,
overrides `.page-header { justify-content: space-between }` |
| `.compare-breadcrumbs` | Themed breadcrumb link strip |
| `.compare-controls`, `.compare-selector`, `.compare-select-group`,
`.compare-select`, `.compare-vs`, `.compare-btn` | Selector strip —
re-themed with surface tokens |
| `.compare-strip`, `.compare-strip-row`, `.compare-strip-side`,
`.compare-strip-mid`, `.compare-strip-name`, `.compare-strip-count`,
`.compare-strip-mid-count`, `.compare-strip-mid-label`,
`.compare-strip-sub` | The headline small-multiples row (A \| shared \|
B) |
| `.compare-bar`, `.compare-bar-seg`, `.compare-bar-{a,both,b}`,
`.compare-bar-legend`, `.compare-legend-item`, `.compare-dot-{a,both,b}`
| Single proportional diff bar |
| `.compare-asym`, `.compare-asym-line`, `.compare-asym-pct` | Compact
directional-reach sentences (replaced the two big cards) |
| `.compare-type-summary`, `.compare-type-summary-label`,
`.compare-type-badge` | Shared-type pill row with ctx-style border-left
accent |
| `.compare-tabs`, `.compare-tabs .tab-btn`, `.tab-btn.active` | Tabs
reskinned to match the muted-then-accent pattern |
| `.compare-summary-text`, `.compare-warning`, `.compare-good` | Themed
status notes |
| `.col-compare-select`, `.col-compare-select input[type="checkbox"]` |
Compare-select column — muted when empty, full text + `--selected-bg`
row tint when populated |
| `.obs-table.has-compare-selection` | Marker class so the column
changes intensity only when something is picked |
| `.observers-page .page-header`, `.obs-refresh-spacer` | Header layout
(flex with right-side refresh icon) |
| `.observer-detail-page .compare-with-group` | Grouped picker + Compare
button surface on the detail page |
**Tokens used:** `--surface-1`, `--surface-2`, `--border`, `--accent`,
`--accent-hover`, `--text`, `--text-muted`, `--row-hover`,
`--hover-bg`, `--selected-bg`, `--status-green`, `--status-amber`,
`--status-amber-light`, `--status-amber-text`, `--radius-sm`,
`--radius-md`, `--badge-radius`, `--space-xs..xl`, `--fs-sm..xl`,
`--mono`. **No new top-level color tokens were introduced.**
## Before
PR #1642's bare `<button class="btn-secondary">` rendered with the
browser-default white pill and the compare page showed three rgba-tinted
cards (`rgba(34,197,94,0.1)`, `rgba(74,158,255,0.1)`,
`rgba(255,107,107,0.1)`) — chartjunk with no theme awareness. See
#1644 description for the bug repro.
## After (screenshots)
**Desktop — observers page (light, empty + selected states):**
- Empty: `MEDIA: 42d90aa5-643c-4e88-8b5d-3383cfa2dfe4.jpg`
- Two selected (rows tinted, button enabled): `MEDIA:
a6d9b397-ffe5-4eeb-b07b-ef89041ab6ea.jpg`
**Desktop — observer detail (light, picker + Compare grouped):**
`MEDIA: 17b9b47d-5e97-4293-8558-e9b37c244335.jpg`
**Desktop — compare page (light, real data via mock — fixture has 0
overlap):**
`MEDIA: be169bf2-f31b-480a-97b1-4f678745471b.jpg`
**Desktop — compare page (dark):**
`MEDIA: 436477a7-600c-4ac4-aa9d-97db968246d3.jpg`
**Desktop — observers (dark, two selected):**
`MEDIA: 850242c3-db77-460f-895f-0a6e6b150758.jpg`
**Mobile 375px — observers (dark):**
`MEDIA: 338b543c-0705-41ec-95da-e2c2a8db2065.jpg`
**Mobile 375px — compare page (dark, stacks cleanly):**
`MEDIA: 380a984c-26f0-4f47-b4ba-d655571721c9.jpg`
## Test plan
- `node test-issue-1644-redesign.js` — 8/8 (new behavioral suite for
this PR)
- `node test-issue-1562-observers-summary.js` — 13/13
- `node test-compare-overlap.js` — 6/6
- `node test-compare-flood-filter.js` — 6/6
- `node test-frontend-helpers.js` — 611/611
- `node scripts/check-css-vars.js` — 0 undefined refs across 1901 var()
calls
- Browser-validated against local fixture build at `localhost:13580`:
desktop light/dark, mobile 375px light/dark, observers + detail +
compare pages. Checkbox preservation verified by manual refresh
click — state survives the tbody rewrite.
## TDD
- Red commit: `94e019c5` — 7 behavioral assertions that all FAIL on
master (no top-level `.btn-secondary`, no `preserveCompareSelection`
helper, rgba literals in compare-card rules).
- Green commit: `a246208d` — implementation. All 8 assertions pass
(the rgba assertion was relaxed to a conditional check after the
cards were removed entirely in favor of the strip; an additional
`.compare-strip exists` assertion was added).
## Out of scope
- The server-side `&since=...` parser is strict about RFC3339 and
rejects the `.000Z` suffix the frontend emits; this means the
comparison page shows zeros against any data > 24h old. Filed
separately — not a regression introduced by this PR. Screenshots
showing populated numbers use a `comparePacketSets` test stub.
- Backend Go untouched.
Fixes#1644
---------
Co-authored-by: clawbot <clawbot@users.noreply.github.com>
Co-authored-by: openclaw-bot <bot@openclaw>
## Summary
The observer-comparison page (`#/compare`) is a powerful side-by-side
overlap tool but was reachable from exactly one place — an icon-only 🔍
button in the observers page header. Most operators never found it. This
PR promotes it to an IA citizen with **three new entry points** plus
breadcrumbs back from the compare page to each observer's detail page.
Red commit: `f937d29658e25973786f88a9ddeaaa33768f269e` (test asserts all
three new affordances are present + navigate correctly; would have
caught the original undiscoverability).
Green commit: `5ceb34b66d780a971d3a43de06a0744445bdbecf`.
## Design rationale
Three orthogonal user paths reach the same goal:
- **Operator who lands on `/observers`** sees a labeled button — no more
icon-guessing — and a row-selection workflow for direct manipulation
("pick two, compare").
- **Operator who lands on a specific observer's page** sees an
in-context "Compare with…" picker — the comparison is parameterised with
the current observer, removing the cognitive jump back to the list.
- **Operator who already has two observer IDs** can still hit
`#/compare?a=…&b=…` directly — legacy deep-links regression-guarded by
the E2E.
Plus: every compare-page view now shows `Observers › <A> ⇆ <B>`
breadcrumbs that link back to each observer's detail page, so users can
navigate sideways instead of bouncing through the list.
## Entry points added
| # | Surface | Affordance | File:line |
|---|---|---|---|
| A | `/observers` header | `<button>` labeled "🔍 Compare observers" |
`public/observers.js:125-130` |
| B | `/observers/<id>` header | "Compare with…" `<select>` + Compare
button | `public/observer-detail.js:90-103`, `:128-145`, `:436-456` |
| D | `/observers` table | Per-row checkbox column + "Compare selected
(N)" button enabled at exactly 2 | `public/observers.js:131-137`,
`:295-302`, `:148-167`, `:354-378` |
| breadcrumbs | `/compare` page | `data-role="compare-breadcrumbs"` with
linked anchors → both detail pages | `public/compare.js:108`, `:202-228`
|
The pre-existing 🔍 link was REMOVED and replaced by (A) — the issue
explicitly called for the icon-only affordance to go away.
## Before — current state on staging
- Observers page header has only a bare 🔍 icon — no text label,
indistinguishable from a generic search affordance.
- Observer-detail page has zero comparison affordances; the user has to
back out, find the observers list, locate the icon, then re-select both
observers from scratch.
- Compare page has a single back-arrow to `/observers` but no breadcrumb
links to either compared observer's detail page.
## After — each new entry point browser-verified locally
Built `cmd/server`, ran against `test-fixtures/e2e-fixture.db` on
`:13581`, drove via headless chromium. Each step taken from a clean
reload, screenshot captured (attached separately to the requesting
session):
- (A) Observers page header now shows a clearly-labeled "🔍 Compare
observers" button alongside a "⚖️ Compare selected (N)" button (disabled
when count !== 2).
- (D) Two rows checked → "Compare selected (2)" enables → click →
navigates to `#/compare?a=…&b=…` with both selects pre-populated and
breadcrumbs reading `Observers › Kennedy Repeater ⇆ GY889 Repeater`.
- (B) Observer-detail header now hosts a "Compare with…" `<select>`
populated with the 30 other observers + a Compare button (disabled until
a target is picked) → pick + click → navigates with the current observer
pre-set as A.
- Legacy `#/compare?a=…&b=…` deep-link still pre-populates both selects
unchanged (covered by the E2E regression guard).
## Test plan
- New: `test-issue-1640-compare-discovery-e2e.js` — 9 assertions across
all three entry points + breadcrumbs + legacy-deep-link regression
guard. Wired into `.github/workflows/deploy.yml`.
- Local browser-verified each new affordance end-to-end (screenshots
above).
- `node --check test-issue-1640-compare-discovery-e2e.js` ✅
- Preflight clean (all 11 gates ✅), see below.
## Preflight checklist
```
── [GATE] PII ── ✅ pass
── [GATE] Branch scope ── ✅ pass (5 files: 1 workflow, 3 frontend, 1 E2E)
── [GATE] Red commit ── ✅ pass (f937d29 verified failing)
── [GATE] CSS-var defined ── ✅ pass
── [GATE] CSS self-fallback ── ✅ pass
── [GATE] LIKE-on-JSON ── ✅ pass
── [GATE] Sync migration ── ✅ pass
── [GATE] Async-migration gate ── ✅ pass
── [GATE] XSS sinks ── ✅ pass
── [WARN] img/SVG ratio ── ✅ pass
── [WARN] Themed <img> SVG ── ✅ pass
── [WARN] Fixture coverage ── ✅ pass
═══ Preflight clean. ═══
```
## Accessibility
- (A) and "Compare selected" buttons carry both visible text AND
`aria-label`; disabled state uses both `disabled` and
`aria-disabled="true"`.
- (B) picker has an `<label class="sr-only">` plus `aria-label` for
screen readers.
- (D) per-row checkbox has `aria-label="Select <observer name> for
comparison"`.
- Breadcrumbs use `<nav aria-label="Compare breadcrumbs">` with a
meaningful `›` separator (aria-hidden).
## Out of scope
- The compare engine itself (`public/compare.js` data flow) is
untouched.
- New comparison metrics (track #671).
- Analytics-nav link suggested as option (C) in the issue — covered by
(A) which is more visible at the same top-nav tier; happy to add later
if needed.
Fixes#1640
---------
Co-authored-by: clawbot <bot@openclaw>
## Summary
Wires the shared `TableSort` helper (already used by the nodes table,
#679) into the observers table at `#/observers`. Adds `data-sort-key` /
`data-type` attrs on every `<th>`, `data-value` on every `<td>` with the
raw sortable value (epoch-ms for times, integers for counts, abs-seconds
for clock skew, derived health rank for the status dot), and initializes
`TableSort` at the end of `render()` — after the new `tbody` is in the
DOM — to avoid the #679 init race on async refresh.
## Before / after
- **Before:** clicking any column header on `#/observers` does nothing —
bare `<th>` cells, no click handlers, no `TableSort.init` call (per
#1639 repro).
- **After:** clicking a header toggles asc/desc with `aria-sort`
indicator + ▲/▼ glyph. Numeric columns (Packet Health, Total Packets,
Packets/Hour, Clock Offset, Uptime) sort numerically. Time columns (Last
Status, Last Packet) sort by ISO timestamp, not the `"23d ago"` display
string. Active column + direction persisted in `localStorage` under
`meshcore-observers-sort`. Default sort: Last Status desc (matches
existing default ordering).
## Test plan
- TDD red commit `0dcd5304` — fails on assertion `Total Packets <th>
must carry data-sort-key="packet_count"` against master.
- Green commit `d4f0376f` — both assertions pass.
- E2E assertion added: `test-issue-1639-observers-sort-e2e.js:46`
(header has `data-sort-key`+`data-type`) and `:62` (click reorders rows
numerically desc).
- Local commands run from the worktree:
- `cd cmd/migrate && go build -o ../../cs-migrate-1639 .` →
`./cs-migrate-1639 -db test-fixtures/e2e-fixture.db`
- `cd cmd/server && go build -o ../../cs-server-1639 .` → run on port
13581 against the fixture DB
- `CHROMIUM_PATH=/usr/bin/chromium BASE_URL=http://localhost:13581 node
test-issue-1639-observers-sort-e2e.js` → ✅ both tests pass
- `node test-observers-headings.js` (#1039 regression) → ✅ still passes
- Browser verified: headless chromium against the local fixture server.
Clicked Total Packets header three times: first click →
`aria-sort=descending` + ▼ glyph + rows ordered 139,261 → 5,791. Second
click → `aria-sort=ascending` + ▲ glyph. Third click → back to
descending. tbody re-renders correctly after the 30s `loadObservers`
auto-refresh (no init race — the new TableSort controller binds to the
fresh header).
- pr-preflight: clean (all hard gates + warnings pass against
`origin/master`).
## Files changed
- `public/observers.js` — wire TableSort, add
`data-sort-key`/`data-type`/`data-value`, init after render
- `test-issue-1639-observers-sort-e2e.js` — new E2E (red→green)
- `.github/workflows/deploy.yml` — run the new E2E alongside existing
playwright group
Fixes#1639
---------
Co-authored-by: openclaw-bot <bot@openclaw>
Co-authored-by: clawbot <clawbot@users.noreply.github.com>
Co-authored-by: openclaw-bot <bot@openclaw.local>
## Summary
The server clamps `/api/nodes` `?limit` to **500** (DoS guard, PR #1540
/ v3.8.3) and orders by `last_seen DESC`. Every node-list consumer
issued a single big-`?limit` fetch and trusted it as the full set, so on
>500-node meshes the top-500-by-advert window silently hid the tail.
Because `nodes.last_seen` is updated **only on self-adverts** (never on
relay traffic; `UpsertNode` is called solely from the advert path), a
repeater that relays constantly but last advertised hours ago fell
outside that window and **vanished from the map and live view** — while
still showing "Active" in its detail panel and (since #1606) in the
paginated Nodes list.
#1606 fixed only the Nodes page (`nodes.js`). This generalizes that fix
to the deferred siblings.
## Changes
- **`public/app.js`** — new shared `fetchAllNodes(extraQuery, opts)`:
pages `limit=500` + `offset` until a short page (the server's `total` is
unreliable — clamped to the page size and overwritten with the filtered
length under area/region filters, so we stop on a short page, not on
`total`), dedups by `public_key`, returns the real deduped count as
`total`.
- **`public/map.js`**, **`public/live.js`** (keeps the
`LIVE_MAP_MAX_NODES` ceiling via `safetyCap`), **`public/analytics.js`**
(×2), **`public/packets.js`** now use the helper.
- **`public/area-map.html`** is standalone (cross-origin `baseUrl`, no
`app.js`) so it gets an inline copy of the same loop.
- **`.eslintrc.json`** — declare `fetchAllNodes` global (no-undef).
## Tests
- **`test-fetch-all-nodes-pagination.js`** — unit-tests the helper via
the real `api()`+`fetch` path: pagination past 500, short-page stop vs.
the unreliable server `total`, dedup across a page boundary, counts
pass-through, `safetyCap` bound. 5/5.
- **`test-map-nodes-pagination-e2e.js`** — browser E2E (Playwright)
proving `map.js` surfaces a 501st node reachable only on page 2 and
renders its marker. Verified **red→green**: against the pre-fix single
fetch all 3 assertions fail (500 nodes, page-2 node absent, no marker);
after the fix all pass. Wired into `deploy.yml`.
## Verification
- unit 5/5, E2E 3/3, `test-frontend-helpers.js` 611/611, `npx eslint
public/*.js` → 0 errors.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Red commit: 178617ca7b6a3b34597ae59d83e15801985be949 (CI run:
https://github.com/Kpa-clawbot/CoreScope/actions/runs/27191921487 —
red-state was verified locally; CI on this branch runs against green
HEAD per pull_request triggers)
Fixes#1629
## Summary
`/api/nodes/{pubkey}/reach` cached responses survived blacklist
mutations for up to the 5-minute TTL. A node added to `NodeBlacklist`
after a recent reach request was still served the cached non-blacklisted
payload until the entry expired.
## Fix (per triage)
Per @Kpa-clawbot's locked fix path on the issue:
1. Add a monotonic `BlacklistGeneration()` counter on `*Config`.
2. `SetNodeBlacklist` (new setter) atomically replaces the slice,
rebuilds the lookup set under an `RWMutex`, and bumps the generation via
`atomic.AddUint64`.
3. `cmd/server/node_reach.go` folds the generation into the cache key
(`"<pubkey>|<days>|g<gen>"`) so any mutation invalidates prior entries
on the next request — no callbacks bolted onto the setter, no
cache-layer surgery, no TTL change.
While here, the latent bug in `blacklistSet()` is also fixed:
`sync.Once` locked in the initial set, so a later `SetNodeBlacklist` was
invisible to `IsBlacklisted`. The `Once` still gates the lock-free
initial build; mutations rebuild under `RWMutex` and reads take an
`RLock` around the map handoff.
## Files
- `cmd/server/config.go` — `SetNodeBlacklist`, `BlacklistGeneration`,
`rebuildBlacklistSetLocked`, `RWMutex`. `IsBlacklisted` reads the
rebuilt set (no stale-slice short-circuit).
- `cmd/server/node_reach.go` — `cacheKey` includes `|g<gen>`.
- `cmd/server/node_reach_blacklist_cache_test.go` — new regression test
(the red commit).
- `cmd/server/node_reach_endpoint_test.go` — existing cache-hit
assertion updated to the generation-suffixed key.
## TDD evidence
- Red commit `178617ca` adds the test + a deliberate `SetNodeBlacklist`
stub that only reassigns the slice. The test fails on the post-blacklist
assertion: `status=200 want 404 (cached payload was served — #1629)`.
- Green commit `257c104f` replaces the stub with the real
implementation; full `go test ./...` and `go test -race -run
"TestNodeReach|TestNodeBlacklist|TestConfig"` pass locally.
## Scope
- One narrow PR. Backend only — no frontend or API response-shape
change.
- No public type signatures touched beyond the new exported
`SetNodeBlacklist` / `BlacklistGeneration` on `*Config`.
- Preflight: all hard gates pass (PII, branch scope, red commit, CSS,
LIKE/JSON, sync/async migration, XSS).
---------
Co-authored-by: corescope-bot <bot@corescope.local>
Co-authored-by: openclaw-bot <bot@openclaw.local>
Red commit: 03546923b4 (CI run: pending —
see Checks)
E2E assertion added: test-issue-1630-reach-mobile-e2e.js:97
## Summary
Adds narrow-viewport CSS to `public/node-reach.css` so the
`/nodes/{pubkey}/reach` page no longer overflows phone-class viewports.
Fixes#1630
## Approach (red → green)
1. **RED** (`03546923`): added `test-issue-1630-reach-mobile-e2e.js`
asserting at 393×800 and 360×740 that:
- `#nqMap` computed height ≤ 320px
- `.nq-table` scrollWidth ≤ clientWidth (no inner h-scroll)
- ≤ 4 visible TH columns (low-signal collapsed)
Desktop guard at 1440×900: map height stays ~420px and all 6 columns
remain visible — proves no desktop regression.
Wired into `.github/workflows/deploy.yml` Playwright job so CI is the
source of truth.
2. **GREEN**: added `@media (max-width: 480px)` block in
`public/node-reach.css` that shrinks `.nq-map` to 280px, hides the
`distance (km)` column, and stacks `we hear` / `they hear us` into a
single compact column.
## Out of scope (intentionally not touched)
- Backend `cmd/server/node_reach.go` (tracked in #1631 / #1629).
- Reach page re-theming.
- Per-column user toggles.
## Local verification
Screenshots at the three target viewports (393×800, 360×740, 1440×900)
attached below.
---------
Co-authored-by: openclaw-bot <bot@openclaw.local>
## TL;DR
Post-merge regression introduced by #1627 r3 (commit `e2212f50`):
`buildNodeInfoMap` in `cmd/server/neighbor_api.go` ran an uncached
`SELECT … FROM nodes` scan on every call. Folded `first_seen` into the
already-cached `getCachedNodesAndPM` (30s TTL) so the 4 hot handlers
that call `buildNodeInfoMap` no longer pay for a full table scan per
request.
## Before / After
`buildNodeInfoMap` is called by **4 hot handlers**:
- `cmd/server/neighbor_api.go:130`
- `cmd/server/neighbor_api.go:297`
- `cmd/server/neighbor_debug.go:83`
- `cmd/server/node_reach.go:421`
| | Before | After |
|---|---|---|
| `SELECT … FROM nodes` per call | 1 (uncached) | 0 (cache hit) |
| `SELECT … FROM observers` per call | 1 (uncached) | 1 (unchanged) |
| At Cascadia scale (~2600 nodes) | full scan × 4 handlers × N req/s |
one scan / 30s |
## How
- Extended the `getAllNodes` schema probe to also `COALESCE(first_seen,
'')`. Falls back through the existing richest → leanest ladder if the
column is missing.
- `nodeInfo.FirstSeen` is therefore populated for every cached entry in
`getCachedNodesAndPM`.
- `buildNodeInfoMap` drops its second `SELECT` entirely and just copies
`nodeInfo` values out of the cached map.
- Public signature of `buildNodeInfoMap` is unchanged.
`node_reach.go:421` still sees `nodeInfo.FirstSeen` populated, served
from cache.
`cmd/server/store.go` is touched because `getAllNodes` is the only
sensible owner of the `first_seen` SELECT — adding a parallel cache
would duplicate the 30s TTL machinery this fix is designed to leverage.
## Test (red → green)
- Commit 1 (`test:`): `TestBuildNodeInfoMap_FirstSeenIsCached` — calls
`buildNodeInfoMap`, mutates `first_seen` out-of-band via a separate rw
connection, calls it again, and asserts both calls return the same
(cached) value. Fails on `origin/master` (call 2 sees the mutated value,
proving the uncached scan).
- Commit 2 (`perf:`): the fold. Test now passes.
## Refs
Post-merge audit identified this as the only MAJOR finding from #1627;
recommendation was a follow-up hot-fix PR. This is that PR.
---------
Co-authored-by: openclaw-bot <bot@openclaw>
Co-authored-by: openclaw-bot <bot@openclaw.local>
