feat(02-11): profile the six regressed surfaces — real cause found before any fix

CPU profile evidence (CDP Profiler + Tracing, additive
neode-ui/e2e/perf/profile-revisit.spec.ts, frozen harness untouched)
shows 86-99% of every revisit window spent in (idle)/(program) with
under 10% genuine app JS self-time on every surface — ruling out
expensive computed re-evaluation, watcher cascades, and whole-subtree
re-renders as the dominant cost, per the plan's own explicit list of
hypotheses to check before accepting one.

A follow-up source-level lifecycle audit (grep every setInterval call
site for a missing onActivated/onDeactivated pair, extending 02-04's
own audit convention past the top-level view files it originally
checked) found three child components/composables inside the
KeepAlive'd Fleet/Server/Web5 subtrees that arm a poll in onMounted
and only ever clear it in onUnmounted — harmless before phase 2
(the view was destroyed on tab-away) and now a permanent, session-long
background-RPC cost once KeepAlive keeps the parent instance alive:
useFleetData.ts (60s), FipsNetworkCard.vue (15s, Server), and
Web5Monitoring.vue (30s, Web5 — redundant with Home.vue's own,
correctly-gated 10s poll of the same store).

This directly explains the idle-dominated CPU signature (background
network/scheduling contention, not compute) and why 02-10's 5-run
remeasure regressed further than the 3-run baseline/after runs even
as disk pressure eased: a longer session accumulates more of these
always-on pollers, degrading every subsequent navigation.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
archipelago 2026-07-31 08:24:59 -04:00
parent 6bf33b57b6
commit 050a87d2dd
2 changed files with 607 additions and 0 deletions

View File

@ -855,3 +855,190 @@ re-measurement.
Wallet/send-flow is **not** in this table — it cleared as noise (see verdict above) and needs no
deviation entry. Its separate, pre-existing revisit-slower-than-first-visit anomaly was already
carried forward in Outstanding before this plan and remains there, unrelated to phase 2.
## Client-Side Render Cost Root Cause (gap closure, 02-11)
**Named cause (one sentence):** the dominant per-revisit cost on all six surfaces is NOT
CPU-bound render/computed work at all (a real CPU profile shows 90-99% of every window spent in
`(idle)`/`(program)`, with genuine app JS self-time under 10% everywhere) — it is background
network/scheduling contention from THREE child components/composables that arm a `setInterval`
poll in `onMounted` and only ever tear it down in `onUnmounted`/`onBeforeUnmount`, a pattern that
was harmless before phase 2 (the view was destroyed on every tab-away, so the poll died with it)
and is now a permanent, session-long background cost once phase 2's KeepAlive wrapping (02-04)
keeps the parent view's instance — and therefore these un-audited children — alive forever.
### Method
Built `neode-ui/e2e/perf/profile-revisit.spec.ts` (additive, does not touch the frozen
`surfaces.ts`/`measure.ts`/`surface-perf.spec.ts` — confirmed via `git diff --stat 3ee20430 --
neode-ui/e2e/perf/surfaces.ts neode-ui/e2e/perf/measure.ts neode-ui/e2e/perf/surface-perf.spec.ts`,
empty both before and after this task). For each of the six named surfaces, reproduces the
harness's own first-visit -> away -> revisit structure, then during the revisit window only,
captures: a CDP `Profiler` CPU sample (self-time by function and by deployed chunk — production
has no sourcemaps, confirmed via `curl -I http://archi-dev-box/assets/vendor-*.js.map` -> 404, so
bucketing is by chunk name not source path), a CDP `Tracing` capture (the same category
`disabled-by-default-devtools.timeline` DevTools' own Performance panel uses), every
`setTimeout`/`requestAnimationFrame` call scheduled during the window (via a runtime monkey-patch,
restored after read-back), every RPC call's wall-clock start/duration, every CSS
`transitionrun`/`transitionend`/`animationstart`/`animationend` event on `document`, and an
independent first-paint timestamp (a rAF-driven poll for `contentSelector`'s first non-empty
bounding box — NOT gated by Playwright's own stricter `visible` actionability rules, which check
only bounding-box-non-empty + not-`visibility:hidden`, never opacity/transform, so a CSS
transition's duration does not by itself block the harness's own measured field).
Ran against `archi-dev-box` (`ARCHY_BASE_URL=http://archi-dev-box`, `ARCHY_PASSWORD` from the
environment). Four of six surfaces completed cleanly in the first pass (Discover, AppDetails,
OpenWrtGateway, Fleet); a second pass added Web5 and Server and the Tracing/animation
instrumentation described above. A later attempt in this same investigation hit the
`OpenWrtGateway` nav step ("waiting for locator('a:has-text(\"OpenWrt Gateway\")').first()") — not
treated as a measurement, per this doc's own standing rule that a harness-execution failure is
never silently written in as data; `dismissOverlays()` was hardened afterward (HealthNotifications'
bare-SVG dismiss button has no `aria-label`, so the original `aria-label*="Close"` selector never
matched it — the exact intercepting-toast risk this doc's own Outstanding section already flagged
for the disk-usage banner) and `clickWithGuard`'s attempt count/force-click fallback increased.
### CPU-profile evidence: the cost is NOT computation, it's waiting
Real, sampled (100us interval) CPU self-time by category, one revisit window per surface (single
representative run each — dispersion comes from Task 3's 5-run re-measure once a fix lands, this
task's job is naming the mechanism):
| Surface | Window (ms) | `(idle)` | `(program)` | All app JS (vendor+index+per-route chunks combined) | rAF calls | setTimeout calls |
|---|---|---|---|---|---|---|
| Web5 | 1452 | 730.5ms (50.4%) | 645.9ms (44.6%) | ~19ms (1.3%) | 12 | 17 (incl. six `15000`/five `180000` — poll re-arms, not blocking) |
| Server | 160 | 90.5ms (56.9%) | 56.4ms (35.4%) | ~1ms (0.6%) | 2 | 3 (all `180000`) |
| Discover | 1436 | 919.4ms (64.0%) | 300.8ms (21.0%) | ~98ms (6.9%) | 44 | 12 (incl. paired `150/150,301/301,501/501`) |
| AppDetails | 1085 | 716.7ms (66.1%) | 217.9ms (20.1%) | ~62ms (5.7%) | 44 | 17 |
| OpenWrtGateway | — | — | — | — | — | — (harness-execution failure this pass, see Method — not treated as data) |
| Fleet | 982 | 706.6ms (72.0%) | 143.1ms (14.6%) | ~48ms (4.9%) | — | — (captured in an earlier pass without the timer instrumentation; re-captured alongside the fix's own before/after in Task 2/3) |
**What this table rules out, directly (not by inference):** every one of the hypotheses this
plan's own objective flagged as plausible — "expensive `computed` re-evaluation over large lists
on reactivation," "watchers cascading on activate," "the whole cached subtree re-rendering,"
"per-row formatting recomputed for every item," "large `v-for` lists without stable
keys/virtualization" — would show up as real, attributed JS self-time inside app-code call frames.
It does not: combined vendor+index+per-route-chunk self-time is 1-7% of the window on every
surface, `(idle)`+`(program)` is 86-99%. The main thread is not busy; it is waiting. This
redirected the investigation from "what render work got heavier" to "what is the browser waiting
on," per D-10's own instruction to add instrumentation rather than assert an unsupported
hypothesis when the profile is ambiguous.
### Source-level lifecycle audit: three permanently-running background pollers, invisible to 02-04's own per-view audit
02-04's lifecycle audit (`02-04-SUMMARY.md`'s per-view side-effect table) grepped each TOP-LEVEL
view file (Home.vue, Web5.vue, Server.vue, Fleet.vue, ...) for timer/subscription/listener tokens
and correctly bucketed each one found into once-per-session / every-entry / only-while-visible.
Fleet.vue itself was checked and correctly found clean ("confirmed no lifecycle side effects (grep
for the five tokens found none)") — but that check never extended to the CHILD components and
composables the view delegates real work to. A repeat of the same grep across every file in
`neode-ui/src` (not just the audited top-level views), cross-checked for `setInterval` present
without a matching `onActivated`/`onDeactivated` pair, found three offenders exactly in the
KeepAlive'd subtrees of three of this gap's six named surfaces:
1. **`neode-ui/src/views/fleet/useFleetData.ts:483-495`** (Fleet.vue's own data composable —
`Fleet.vue` calls `useFleetData()` directly, `const fleet = useFleetData()`):
```
onMounted(async () => {
...
await refreshAll()
if (autoRefresh.value) startAutoRefresh() // pollTimer = setInterval(..., 60000)
})
onUnmounted(() => { stopAutoRefresh() }) // never fires again once Fleet is KeepAlive'd
```
`startAutoRefresh()`'s `pollTimer` fires `Promise.all([fetchFleetStatus(), fetchFleetAlerts()])`
(plus `fetchNodeHistory()` if a node is selected) every 60 seconds, forever, from the first
Fleet visit onward, regardless of whether Fleet is the visible dashboard tab.
2. **`neode-ui/src/views/server/FipsNetworkCard.vue:265-274`** (rendered inside `Server.vue` at
line 164, `<FipsNetworkCard />`):
```
let statusInterval: ReturnType<typeof setInterval> | null = null
onMounted(() => {
statusInterval = setInterval(() => { if (document.hidden) return; void statusRes.refresh() }, 15000)
})
onUnmounted(() => { if (statusInterval) clearInterval(statusInterval) })
```
The `document.hidden` guard only helps when the whole BROWSER TAB is backgrounded — it does
nothing for an in-SPA dashboard-tab switch, since `document.hidden` stays `false` the entire
time the Archipelago tab itself is foregrounded, regardless of which dashboard view is showing.
This fires an `fips.status` refresh every 15 seconds, forever, from the first Server visit
onward, once Server is KeepAlive'd.
3. **`neode-ui/src/views/web5/Web5Monitoring.vue:105-114`** (rendered inside `Web5.vue` at line
72, `<Web5Monitoring />`):
```
onMounted(() => { loadStats(); refreshInterval = setInterval(loadStats, 30000) })
onBeforeUnmount(() => { if (refreshInterval) clearInterval(refreshInterval) })
```
`loadStats()` calls `homeStatus.refreshSystemStats()` — the SAME store Home.vue's own,
correctly `onActivated`/`onDeactivated`-gated 10s poll already refreshes (02-04's own table:
`Home.vue | systemStatsInterval (10s loadSystemStats) | V`). This is a second, redundant,
never-torn-down 30s poll of the identical data, forever, from the first Web5 visit onward.
Confirmed (not assumed) that this is a genuine phase-2 regression, not pre-existing: before 02-04
registered these three views in `KEEP_ALIVE_PATHS`, leaving any of them fully **unmounted** the
component (and every child, including these three), which fired `onUnmounted`/`onBeforeUnmount`
and cleared the interval — the exact class of bug 02-04's audit convention exists to prevent,
missed here only because the audit's own grep scope was the view file, not its full child tree.
Every other main-tab-adjacent file with a `setInterval` was checked the same way
(`grep -rl setInterval src/ | xargs check-for-onActivated`) — `Home.vue`, `Web5.vue` (own poll),
`Server.vue` (own poll) and `Mesh.vue` all correctly gate their OWN top-level intervals (confirmed
`onActivated`/`onDeactivated` present per 02-04's table); the three offenders above are the only
KeepAlive-subtree-reachable files missing the pair. (`useControllerNav.ts`'s route-change
`setTimeout(autoFocusMain, 150/300)` was also checked and ruled out — `autoFocusMain()` early-
returns unless `navRingActive()`, which is `false` for pointer/mouse-driven navigation; confirmed
negligible self-time in every profile above, so it does not explain a mouse-driven revisit's wall
clock, only a gamepad/keyboard user's.)
### Why this produces exactly the `(idle)`-dominated signature measured above
Three RPC-driven pollers running unconditionally every 15-60 seconds for the rest of the session,
completely independent of which dashboard tab is in the foreground, is real background network and
main-thread-scheduling activity that a JS CPU profile attributes to `(idle)`/`(program)` (waiting
on the network stack, connection-pool bookkeeping, and scheduler turnarounds) rather than to any
named app function — which is exactly what the table above shows, and is the same connection-pool-
contention mechanism 02-08 already root-caused for `Cloud.vue`'s peer-browse fan-out (`02-FINDINGS.md`
addendum), just distributed across three smaller, always-on pollers instead of one large burst.
This also explains, for the first time with a mechanism rather than a shrug, why 02-10's
three-way re-measure found the regression getting WORSE across longer/later test runs even as disk
pressure genuinely eased (the opposite of what an environmental-noise theory predicts): a longer
session visits more of Fleet/Server/Web5 at least once, accumulating more of these permanently-
running background pollers, so EVERY subsequent navigation — not just these three surfaces' own
revisits — competes with more background traffic as the session goes on. `ARCHY_PERF_RUNS=5`
(02-10's re-measure) accumulates strictly more of this background load over its ~8-minute run than
`ARCHY_PERF_RUNS=3` (baseline/after) did over a shorter one, independent of any single surface's own
defect.
### Surfaces without their own leaked poller: Discover, AppDetails, OpenWrtGateway
- **OpenWrtGateway**'s `navSteps` transit `/dashboard/server` directly (`surfaces.ts`) — its
measured window already includes Server's own reactivation, so FipsNetworkCard's leaked poller
(and whatever contention it causes) bleeds directly into OpenWrtGateway's number, the same
transit-confound this doc's `## Method` section already documented for its RPC column.
- **Discover** and **AppDetails** both transit `/dashboard/apps` (`surfaces.ts`), and a repeat of
the same lifecycle-audit grep across every child component `Apps.vue` renders
(`AppCard.vue`, `AppIconGrid.vue`, `AppsUninstallModal.vue`) found no `setInterval` at all —
Apps.vue's own audit (02-04) is clean and stays clean here. Their regression is not explained by
a leaked poller of their own. The working hypothesis, to be confirmed (not assumed) by Task 3's
before/after re-measure: session-wide background contention from the three leaked pollers above
degrades ANY foreground navigation once Fleet/Server/Web5 have each been visited at least once in
the session — including Discover/AppDetails, which have no defect of their own to fix. If Task
3's re-measure shows Discover/AppDetails improve materially after fixing only the three pollers,
that confirms this shared-contention mechanism; if they do not move, a surface-specific cause
(most plausibly the `useCachedResource` keyed-cache wrapper's own per-mount setup cost on
AppDetails, which is NOT KeepAlive'd and fully remounts every visit, exceeding what the eliminated
fetch saved — 02-10's own prior "split-signal" framing) remains the standing explanation and is
reported honestly as unresolved by this plan's fix, not glossed over.
### Fix, per this named cause (Task 2)
Apply the exact `onActivated`/`onDeactivated` arm/disarm pattern 02-04 already established and
proved safe for every OTHER poll interval in this codebase (`Server.vue`'s own `vpnPollInterval`,
`Mesh.vue`'s polls, `Home.vue`'s `systemStatsInterval`) to the three files above: start the
interval in `onActivated` (with an immediate first tick, matching the existing convention) and
clear it in `onDeactivated`, keeping `onMounted`/`onUnmounted` as the dual-registration fallback for
a bare (non-KeepAlive) mount, per 02-04's own documented "every arm function is called from BOTH
onMounted and onActivated" rule (a KeepAlive-wrapped first mount fires both once; a fresh-mount
guard is not needed here since the actions themselves — arming a 15/30/60s interval — are cheap and
idempotent, unlike Home/Web5/Mesh/Server's heavier first-load RPC bursts that needed a guard to
avoid double-firing).

View File

@ -0,0 +1,420 @@
// profile-revisit.spec.ts — 02-11 gap-closure diagnostic (D-10: measure before
// fix). Standalone, additive script — does NOT edit surfaces.ts / measure.ts
// / surface-perf.spec.ts (the frozen 02-01 harness stays frozen).
//
// 02-VERIFICATION.md's gap 2 confirmed six surfaces regressed on revisit with
// flat-or-improved RPC counts — i.e. the cost is client-side render/reactivity,
// not network. This script captures a REAL CPU profile (CDP Profiler domain —
// the same sampling data Chrome DevTools' Performance panel visualizes as a
// flame chart) during exactly the harness's own measured window (click chain
// -> contentSelector visible) for each named surface's revisit, then
// aggregates self-time by function so the dominant cost can be named with
// profiling evidence instead of guessed from source reading alone.
//
// Usage:
// ARCHY_BASE_URL=http://archi-dev-box ARCHY_PASSWORD=*** \
// npx playwright test e2e/perf/profile-revisit.spec.ts --project=chromium --reporter=line
import { expect, test, type Page } from '@playwright/test'
import { SURFACES, type Surface } from './surfaces'
const PASSWORD = process.env.ARCHY_PASSWORD ?? 'password123'
const NEUTRAL_SELECTOR = '[data-controller-zone="sidebar"] a[href="/dashboard/settings"]'
const NAV_TIMEOUT = 20_000
const CONTENT_TIMEOUT = 20_000
// Every surface 02-VERIFICATION.md named as regressed, plus Fleet (the
// out-of-scope bonus 02-10 flagged as the most severe magnitude of the same
// mechanism) — all six of this gap plan's must-haves.
const TARGET_IDS = ['web5', 'server', 'discover', 'app-details', 'openwrt-gateway', 'fleet']
async function login(page: Page): Promise<void> {
await page.goto('/login', { waitUntil: 'domcontentloaded' })
await page.evaluate(() => {
localStorage.setItem('neode_intro_seen', '1')
localStorage.setItem('neode_onboarding_complete', '1')
})
await page.goto('/login', { waitUntil: 'networkidle' })
const passwordInput = page.locator('input[type="password"]').first()
await passwordInput.waitFor({ timeout: 15_000 })
await passwordInput.fill(PASSWORD)
await page
.locator('button:has-text("Login"), button:has-text("Unlock"), button:has-text("Continue"), button[type="submit"]')
.first()
.click()
await page.waitForURL('**/dashboard**', { timeout: 20_000 })
}
async function dismissOverlays(page: Page): Promise<void> {
try {
await page.keyboard.press('Escape')
} catch {
/* best-effort */
}
const closeButtons = page.locator(
'[role="dialog"] button[aria-label*="Close" i], .fixed.inset-0 button[aria-label*="Close" i]'
)
const count = await closeButtons.count().catch(() => 0)
if (count > 0) await closeButtons.first().click({ timeout: 2_000 }).catch(() => {})
// HealthNotifications.vue's dismiss button has no aria-label at all (a bare
// SVG X icon) — the aria-label selector above never matches it, and per
// 02-FINDINGS.md's Outstanding section its `.fixed.right-4.z-[200]` wrapper
// has no `pointer-events: none`, so a still-open toast can intercept a
// click meant for page content underneath it (e.g. a disk-usage warning
// sitting over the "OpenWrt Gateway" link). Dismiss any visible one.
const healthToastClose = page.locator('.fixed.right-4.z-\\[200\\] button')
const healthCount = await healthToastClose.count().catch(() => 0)
for (let i = 0; i < healthCount; i++) {
await healthToastClose.first().click({ timeout: 1_000 }).catch(() => {})
}
}
async function clickWithGuard(page: Page, selector: string, timeoutMs: number): Promise<void> {
const attempts = 4
const perAttemptMs = Math.max(2_000, Math.floor(timeoutMs / attempts))
let lastErr: unknown
for (let i = 0; i < attempts; i++) {
await dismissOverlays(page)
try {
await page.locator(selector).first().click({ timeout: perAttemptMs, force: i === attempts - 1 })
return
} catch (err) {
lastErr = err
}
}
throw lastErr instanceof Error ? lastErr : new Error(String(lastErr))
}
async function clickChain(page: Page, steps: string[], timeoutMs: number): Promise<void> {
for (const selector of steps) await clickWithGuard(page, selector, timeoutMs)
}
async function goHome(page: Page, timeoutMs: number): Promise<void> {
if (new URL(page.url()).pathname === '/dashboard/chat') {
const closed = await page.locator('.chat-close-btn').first().click({ timeout: 5_000 }).then(() => true).catch(() => false)
if (!closed) await page.goto('/dashboard', { waitUntil: 'domcontentloaded' }).catch(() => {})
}
await clickWithGuard(page, '[data-controller-zone="sidebar"] a[href="/dashboard"]', timeoutMs)
await page.waitForURL((url) => url.pathname === '/dashboard', { timeout: timeoutMs })
}
interface CpuProfileNode {
id: number
callFrame: { functionName: string; url: string; lineNumber: number; columnNumber: number; scriptId: string }
hitCount?: number
children?: number[]
}
interface CpuProfile {
nodes: CpuProfileNode[]
startTime: number
endTime: number
samples?: number[]
timeDeltas?: number[]
}
// Production build has no sourcemaps deployed (confirmed: assets/*.js.map ->
// 404 on archi-dev-box), so minified function names inside vendor-*.js/
// index-*.js can't be resolved to source. Bucket by the DEPLOYED CHUNK NAME
// instead (still meaningful: vendor = Vue/Pinia/vue-router runtime bundled
// together; index = app entry/shared code; per-route chunk name = that
// view's own lazy-loaded code) rather than by node_modules path, which only
// exists pre-build.
function categorize(url: string): string {
if (!url) return '(native/gc/idle — no JS frame)'
const file = url.split('/').pop() ?? url
if (/^vendor-/.test(file)) return 'assets/vendor-*.js (Vue/Pinia/vue-router runtime bundle)'
if (/^index-/.test(file)) return 'assets/index-*.js (app entry/shared chunk)'
if (/^Dashboard-/.test(file)) return 'assets/Dashboard-*.js (Dashboard/DashboardRouterView chunk)'
if (/^Fleet-/.test(file)) return 'assets/Fleet-*.js (Fleet.vue chunk)'
if (/^Web5-/i.test(file)) return 'assets/Web5-*.js chunk'
if (/^Server-/.test(file)) return 'assets/Server-*.js chunk'
if (/^Discover-/.test(file)) return 'assets/Discover-*.js chunk'
if (/^AppDetails-/.test(file)) return 'assets/AppDetails-*.js chunk'
if (/^OpenWrtGateway-/.test(file)) return 'assets/OpenWrtGateway-*.js chunk'
if (url.startsWith('assets/')) return `assets/${file} (other chunk)`
if (url.includes('node_modules')) return 'other node_modules (dev-only build)'
return '(native/gc/idle — no JS frame)'
}
function analyzeProfile(profile: CpuProfile, label: string): void {
const byId = new Map<number, CpuProfileNode>()
for (const n of profile.nodes) byId.set(n.id, n)
const selfTimeById = new Map<number, number>()
const samples = profile.samples ?? []
const timeDeltas = profile.timeDeltas ?? []
let total = 0
for (let i = 0; i < samples.length; i++) {
const dt = timeDeltas[i] ?? 0
total += dt
const id = samples[i]!
selfTimeById.set(id, (selfTimeById.get(id) ?? 0) + dt)
}
const totalMs = total / 1000
const windowMs = (profile.endTime - profile.startTime) / 1000
// Aggregate by function identity (name@file:line) and by coarse category.
const byFunction = new Map<string, number>()
const byCategory = new Map<string, number>()
for (const [id, us] of selfTimeById) {
const node = byId.get(id)
if (!node) continue
const fnName = node.callFrame.functionName || '(anonymous)'
const file = node.callFrame.url ? node.callFrame.url.split('/').slice(-2).join('/') : '(native)'
const key = `${fnName} @ ${file}:${node.callFrame.lineNumber}`
byFunction.set(key, (byFunction.get(key) ?? 0) + us)
const cat = categorize(node.callFrame.url)
byCategory.set(cat, (byCategory.get(cat) ?? 0) + us)
}
console.log(`\n===== CPU PROFILE: ${label} =====`)
console.log(`Profiled window (Profiler start->stop): ${windowMs.toFixed(1)}ms`)
console.log(`Total sampled self-time: ${totalMs.toFixed(1)}ms (${samples.length} samples)`)
console.log(`--- By category (self time) ---`)
const catSorted = Array.from(byCategory.entries()).sort((a, b) => b[1] - a[1])
for (const [cat, us] of catSorted) {
const ms = us / 1000
console.log(` ${ms.toFixed(1)}ms (${((us / total) * 100).toFixed(1)}%) ${cat}`)
}
console.log(`--- Top 20 functions (self time) ---`)
const fnSorted = Array.from(byFunction.entries()).sort((a, b) => b[1] - a[1]).slice(0, 20)
for (const [key, us] of fnSorted) {
const ms = us / 1000
console.log(` ${ms.toFixed(2)}ms (${((us / total) * 100).toFixed(1)}%) ${key}`)
}
}
test.describe.configure({ mode: 'serial' })
test('profile revisit CPU cost for named regressed surfaces', async ({ page }) => {
test.setTimeout(15 * 60 * 1000)
await login(page)
for (const id of TARGET_IDS) {
const surface = SURFACES.find((s) => s.id === id) as Surface
expect(surface, `surface ${id} must exist in SURFACES`).toBeTruthy()
await goHome(page, NAV_TIMEOUT)
// First visit — warm the cache (matches the harness's own first-visit/
// revisit structure) — not profiled.
await clickChain(page, surface.navSteps, NAV_TIMEOUT)
await page.locator(surface.contentSelector).first().waitFor({ state: 'visible', timeout: CONTENT_TIMEOUT })
// Away — to the neutral Settings tab (or close, for an in-page trigger).
if (surface.closeSelector) {
await page.locator(surface.closeSelector).first().click({ timeout: NAV_TIMEOUT })
} else {
await clickWithGuard(page, NEUTRAL_SELECTOR, NAV_TIMEOUT)
await page.waitForURL((url) => url.pathname === '/dashboard/settings', { timeout: NAV_TIMEOUT })
}
// Let the away-transition settle before starting the profiler so it
// captures only the revisit window, not Settings' own leave-transition.
await page.waitForTimeout(600)
// Instrument window.setTimeout/requestAnimationFrame for the profiled
// window only, so a large "idle" share in the CPU profile can be
// attributed to a concrete scheduled delay (app timer) vs. a chain of
// animation frames (CSS-transition-bound) vs. neither (network wait).
await page.evaluate(() => {
const w = window as unknown as {
__timerLog: Array<{ type: string; delay?: number; at: number }>
__origSetTimeout: typeof setTimeout
__origRaf: typeof requestAnimationFrame
}
w.__timerLog = []
w.__origSetTimeout = window.setTimeout.bind(window)
w.__origRaf = window.requestAnimationFrame.bind(window)
// eslint-disable-next-line @typescript-eslint/no-explicit-any
;(window as any).setTimeout = (fn: TimerHandler, delay?: number, ...args: unknown[]) => {
w.__timerLog.push({ type: 'setTimeout', delay, at: performance.now() })
return w.__origSetTimeout(fn as any, delay, ...args) // eslint-disable-line @typescript-eslint/no-explicit-any
}
// eslint-disable-next-line @typescript-eslint/no-explicit-any
;(window as any).requestAnimationFrame = (cb: FrameRequestCallback) => {
w.__timerLog.push({ type: 'raf', at: performance.now() })
return w.__origRaf(cb)
}
})
// Wall-clock RPC start/finish (matches measure.ts's own attachRpcTracker
// convention: Date.now() offsets, method name only, no bodies — T-02-06).
const rpcCalls: Array<{ method: string; startedAtMs: number; durationMs: number | null }> = []
const pendingRpc = new Map<import('@playwright/test').Request, { method: string; start: number }>()
let tRpc0 = 0
const isRpc = (req: import('@playwright/test').Request) => req.method() === 'POST' && req.url().includes('/rpc/v1')
const onReq = (req: import('@playwright/test').Request) => {
if (!isRpc(req)) return
let method = 'unknown'
try {
const body = req.postData()
if (body) method = (JSON.parse(body) as { method?: string }).method ?? 'unknown'
} catch {
/* keep 'unknown' */
}
pendingRpc.set(req, { method, start: Date.now() - tRpc0 })
}
const onSettled = (req: import('@playwright/test').Request) => {
const info = pendingRpc.get(req)
if (!info) return
pendingRpc.delete(req)
rpcCalls.push({ method: info.method, startedAtMs: info.start, durationMs: Date.now() - tRpc0 - info.start })
}
page.on('request', onReq)
page.on('requestfinished', onSettled)
page.on('requestfailed', onSettled)
// Capture every CSS transition/animation start+end on the document during
// the window — decisive evidence for/against "a long CSS transition is
// what the human waits through" (Playwright's own 'visible' check does
// NOT wait for transitions/opacity, only a non-empty bounding box + not
// visibility:hidden, so a slow transition would NOT show up as CPU cost
// or as a blocked contentSelector — it would show up here, and would
// explain a "feels slow" gap between first-paint and contentSelector).
await page.evaluate(() => {
const w = window as unknown as { __animLog: Array<{ type: string; name: string; target: string; elapsedMs: number; at: number }> }
w.__animLog = []
const describe = (el: EventTarget | null): string => {
const e = el as HTMLElement | null
if (!e || !e.tagName) return '(unknown)'
const cls = (e.className && typeof e.className === 'string') ? '.' + e.className.trim().split(/\s+/).slice(0, 2).join('.') : ''
return `${e.tagName.toLowerCase()}${cls}`
}
const log = (type: string) => (ev: Event) => {
const te = ev as TransitionEvent | AnimationEvent
w.__animLog.push({
type,
name: (te as TransitionEvent).propertyName ?? (te as AnimationEvent).animationName ?? '',
target: describe(ev.target),
elapsedMs: Math.round((te.elapsedTime ?? 0) * 1000),
at: performance.now(),
})
}
document.addEventListener('transitionrun', log('transitionrun'), true)
document.addEventListener('transitionend', log('transitionend'), true)
document.addEventListener('transitioncancel', log('transitioncancel'), true)
document.addEventListener('animationstart', log('animationstart'), true)
document.addEventListener('animationend', log('animationend'), true)
})
const client = await page.context().newCDPSession(page)
await client.send('Profiler.enable')
await client.send('Profiler.setSamplingInterval', { interval: 100 })
await client.send('Profiler.start')
// Raw Chrome trace events (the SAME data DevTools' Performance panel
// renders as a flame chart / summary tab) — categorizes rendering work
// (Layout, RecalculateStyles, Paint, CompositeLayers, RunTask, TimerFire,
// FireAnimationFrame, ...) that a bare JS CPU profile only sees as
// "(program)"/"(idle)" because layout/paint/compositing run on the same
// renderer main thread but outside any JS call frame.
// eslint-disable-next-line @typescript-eslint/no-explicit-any
const traceEvents: any[] = []
// eslint-disable-next-line @typescript-eslint/no-explicit-any
client.on('Tracing.dataCollected', (data: any) => { traceEvents.push(...(data.value ?? [])) })
const tracingComplete = new Promise<void>((resolve) => client.once('Tracing.tracingComplete', () => resolve()))
await client.send('Tracing.start', {
categories: 'disabled-by-default-devtools.timeline,devtools.timeline,toplevel,v8,blink.user_timing',
transferMode: 'ReportEvents',
})
// First-paint probe, independent of Playwright's own contentSelector
// wait: records performance.now() the moment contentSelector FIRST gets
// a non-empty bounding box, polled every animation frame (not gated by
// Playwright's stricter "visible" actionability rules) — so first-paint
// and content-visible can be reported as two distinct numbers per the
// gap plan's explicit ask.
const pageT0 = await page.evaluate((sel) => {
const w = window as unknown as { __firstPaintAt: number | null }
w.__firstPaintAt = null
function poll() {
if (w.__firstPaintAt == null) {
const el = document.querySelector(sel)
if (el) {
const r = el.getBoundingClientRect()
if (r.width > 0 && r.height > 0) {
w.__firstPaintAt = performance.now()
return
}
}
requestAnimationFrame(poll)
}
}
requestAnimationFrame(poll)
return performance.now()
}, surface.contentSelector)
const t0 = Date.now()
tRpc0 = t0
await clickChain(page, surface.navSteps, NAV_TIMEOUT)
await page.locator(surface.contentSelector).first().waitFor({ state: 'visible', timeout: CONTENT_TIMEOUT })
const wallMs = Date.now() - t0
const firstPaintAt = await page.evaluate(() => (window as unknown as { __firstPaintAt: number | null }).__firstPaintAt)
const firstPaintMs = firstPaintAt != null ? Math.round(firstPaintAt - pageT0) : null
const { profile } = await client.send('Profiler.stop')
await client.send('Profiler.disable')
await client.send('Tracing.end')
await tracingComplete
await client.detach().catch(() => {})
page.off('request', onReq)
page.off('requestfinished', onSettled)
page.off('requestfailed', onSettled)
// Find the renderer main-thread tid (thread_name metadata event) so the
// breakdown below isn't polluted by compositor/IO/GPU-process threads.
const mainThreadMeta = traceEvents.find(
(e) => e.ph === 'M' && e.name === 'thread_name' && (e.args?.name === 'CrRendererMain')
)
const mainTid = mainThreadMeta?.tid
const durByName = new Map<string, number>()
for (const e of traceEvents) {
if (mainTid != null && e.tid !== mainTid) continue
if (e.ph !== 'X') continue // complete events only (have a real duration)
if (typeof e.dur !== 'number') continue
durByName.set(e.name, (durByName.get(e.name) ?? 0) + e.dur)
}
console.log(`--- Chrome trace event self/total time by name (main thread, category=devtools.timeline; NOTE: 'X' events can nest, so this is TOTAL not self time — use for relative magnitude, not a sum-to-100% budget) ---`)
const traceSorted = Array.from(durByName.entries()).sort((a, b) => b[1] - a[1]).slice(0, 20)
for (const [name, us] of traceSorted) {
console.log(` ${(us / 1000).toFixed(1)}ms ${name}`)
}
const timerLog = await page.evaluate(() => {
const w = window as unknown as {
__timerLog: Array<{ type: string; delay?: number; at: number }>
__origSetTimeout: typeof setTimeout
__origRaf: typeof requestAnimationFrame
}
const log = w.__timerLog ?? []
window.setTimeout = w.__origSetTimeout
window.requestAnimationFrame = w.__origRaf
return log
})
const animLog = await page.evaluate(() => (window as unknown as { __animLog: Array<{ type: string; name: string; target: string; elapsedMs: number; at: number }> }).__animLog ?? [])
console.log(`\n>>> ${surface.label} (${surface.id}) revisit wall-clock: ${wallMs}ms | first-paint: ${firstPaintMs}ms (contentSelector's first non-empty bounding box, unrelated to Playwright's own stricter 'visible' check)`)
analyzeProfile(profile as CpuProfile, `${surface.label} (${surface.id})`)
const setTimeoutEntries = timerLog.filter((e) => e.type === 'setTimeout')
const rafCount = timerLog.filter((e) => e.type === 'raf').length
console.log(`--- Timers scheduled during the revisit window (app-code setTimeout/rAF calls) ---`)
console.log(` setTimeout calls: ${setTimeoutEntries.length}${setTimeoutEntries.length ? ' — delays(ms): ' + setTimeoutEntries.map((e) => e.delay ?? 0).sort((a, b) => a - b).join(',') : ''}`)
console.log(` requestAnimationFrame calls: ${rafCount}`)
console.log(`--- CSS transition/animation events during the window (relative to window start, pageT0) ---`)
if (animLog.length === 0) {
console.log(' (none observed)')
} else {
for (const e of animLog) {
console.log(` +${Math.round(e.at - pageT0)}ms ${e.type} name="${e.name}" target=${e.target} elapsed=${e.elapsedMs}ms`)
}
}
console.log(`--- RPC calls during the revisit window (wall-clock ms from window start) ---`)
for (const c of rpcCalls) {
console.log(` ${c.method}: start=+${c.startedAtMs}ms duration=${c.durationMs}ms`)
}
console.log(` total RPC calls: ${rpcCalls.length}, wall-clock window: ${wallMs}ms`)
}
})