test: cover MetricsDashboard's single-point Sparkline arm end to end

[hivecommons-hive]

Test Improvement

Covers the values.length === 1 ? 50 arm of the x-coordinate ternary in
Sparkline (src/components/MetricsDashboard/index.js line 28) — the arm that
centres a lone data point, because the expression beside it divides by
values.length - 1 and would be a division by zero.

No shape in the checked-in data reaches it. data/metrics.json defines exactly
two lifecycle trends: submissions carries zero values, so Sparkline returns
No trend data yet before reaching the ternary at all, and publications
carries five, which takes the arm beside it.

Why the variant build

tests/e2e/fixtures/data/** overlays are additive — the extra record renders
alongside the real ones and nothing already covered stops rendering. Appending
a point to submissions in the ordinary coverage build is not additive in that
sense: submissions is the only trend that reaches the empty-values arm, so
giving it a point would trade that arm for this one rather than add a case.

The point is appended in tests/e2e/fixtures/data-variants/metrics.json
instead, the mechanism tests/e2e/data-variants.spec.js already documents for
this class of branch. One Playwright run then visits /metrics and
/e2e-coverage-variant/metrics and the report unions what each build reached,
so both arms end up covered.

What this adds

• tests/e2e/fixtures/data-variants/metrics.json — one appended point on
  referenceArchitectureLifecycle.trends.submissions.values, beside the
  existing generatedAt clearing. The overlay's append path must already
  exist in the real file, so a regenerated data/metrics.json that drops the
  trend fails the build rather than silently losing the coverage.
• tests/e2e/metrics-sparkline-variant.spec.js — two cases, skipped outside
  E2E_COVERAGE=1. The real route asserts the premise (one trend empty, none
  single-point, No trend data yet rendered); the variant route asserts the
  lone point is centred. Both read the trends through loadSiteData() rather
  than hard-coding a label or a number, so an edited overlay fails here instead
  of leaving a test that asserts nothing. The variant case also asserts the
  other trend still has more than one point, so the variant does not quietly
  trade one uncovered region for another.

No change to any file under src/.

Evidence

• E2E, before — CI artifact e2e-coverage id 11348560988, run
  37317061554,
  job End-to-end coverage, main at 8afaa67 (this branch's base):
  {"file":"src/components/MetricsDashboard/index.js","regions":44,"coveredRegions":38,"regionPercent":86.36,"uncoveredRegions":[28,65,76,85,124,174]}.

• Rendered output, after — npm run build:e2e:coverage:variant at this
  branch's head, then reading the two built pages. Inside
  section[aria-labelledby="lifecycle-title"]:

  build	No trend data yet	sparkline labels	polyline points
  build/metrics (real)	present	Trend from 2024-11 to 2026-07	0,4 25,… 100,…
  build/e2e-coverage-variant/metrics	absent	Trend from 2026-01 to 2026-01, Trend from 2024-11 to 2026-07	50,4, 0,4 25,… 100,…

  points="50,4" is the arm under test: a single pair, x at the centre of the
  0 0 100 24 viewBox rather than an index-derived offset. It is what the new
  spec asserts, computed from the overlaid value rather than written as a
  literal.

• npm run test:unit:coverage:check exits 0 — the overlay change leaves
  tests/e2e-data-fixtures.test.mjs passing, including
  the variant build clears the fields the ordinary build keeps, since
  generatedAt is still cleared to null.

• npx prettier --check clean on both changed files.

• npx playwright test --list collects both cases.

Playwright itself was not run locally: this environment has no root and
chrome-headless-shell cannot load libglib-2.0.so.0, so the browser will not
launch here. The assertions were instead verified against the SSR markup of the
two builds above, which is the same DOM the hydrated page asserts on. CI's
End-to-end coverage job is the authority on the resulting region count.

Related Issue

Closes #1092

Coordination

Touches exactly two files — tests/e2e/fixtures/data-variants/metrics.json and
the new tests/e2e/metrics-sparkline-variant.spec.js — and one function under
test, Sparkline in src/components/MetricsDashboard/index.js.

Disjoint from every open hold-gated PR. In particular it does not edit
tests/e2e/data-variants.spec.js or tests/e2e-data-fixtures.test.mjs, where
#1034 and #1078 are working, which is why the cases live in a new spec file
rather than beside the existing variant-build cases; #1034's variant overlay is
community-people.json, a different file in the same directory. It does not
touch tests/e2e/data-fixtures.spec.js (#1054), the e2e reporter cluster
(#1040, #1070), tests/svg-active-content.test.mjs (#1088, #1090),
scripts/fetch-community-people.mjs (#1085), or the component specs #1058,
#1060, #1065, #1075, #1081, #1083.

The --check-source-regions floor in .github/workflows/ci.yml is not
re-derived here: this change only raises the covered-region count, so the
existing floor still holds, and that path cannot be pushed by this lane in any
case.

— hive: agent=quality backend=copilot model=claude-opus-5 copilot=1.0.88

[hivecommons-hive]

Important

Held for human review by the hive's ACMM level gate.

This PR was opened by the "quality" agent while Hive policy required a human checkpoint for that agent. Non-outreach agents are held at ACMM L3–L5; the outreach agent is always held because it publishes project-facing communication.

Hive will keep the hold label until a human removes it. Operators can make a deliberate one-off release during an ACMM level change with release_level_holds=true, but level changes never release this hold automatically.

[hivecommons-hive]

CI note: runs 37363814397 (Validate repository) and 37363814525 (CodeQL) on the current head (bd7842d) report failure, but no step actually failed: the cancelled jobs are the non-required End-to-end coverage job (cancelled ~15 min in, 19:30→19:45Z) and CodeQL's Analyze JavaScript. Lint, Validate, and End-to-end tests all passed. No replacement run is queued. Same cancellation pattern as #1090 (run 37338929390).

The agent tier cannot re-run jobs (gh run rerun is not allowlisted), so this needs a maintainer: gh run rerun 37363814525 --repo cncf/endusers --failed (and 37363814397) from the Actions UI or CLI. No branch push is needed or appropriate — the diff is not implicated.

---

🐝 Hive Agent: ci-maintainer | Instance: hosted-available-lke648397-260827-5n31 | SHA: unknown

— hive: agent=ci-maintainer backend=copilot model=kimi-k3 copilot=1.0.88