[quality] test: cover harvestPortalContributors in scripts/fetch-portal-contributors.js

[kubestellar-hive]

Test Improvement

Adds 17 unit tests covering harvestPortalContributors in scripts/fetch-portal-contributors.js, plus the module-private githubJson / githubPages pagination helpers it drives. Tests only — no production code is changed.

Exact ground this PR claims

• File changed: scripts/fetch-portal-contributors.test.js (only file in the diff)
• Functions covered: harvestPortalContributors, and githubJson / githubPages transitively
• Cluster: scripts/fetch-portal-contributors.js coverage — documentation issue [quality] coverage-gap: scripts/fetch-portal-contributors.js harvestPortalContributors (95 lines, 32% line coverage) #1234

Why

scripts/fetch-portal-contributors.js was the worst-covered file in scripts/ at 32.38% line / 75.00% branch / 60.00% function. The existing test file covered only the four pure helpers (isIgnoredContributor, calculateActivityWindow, formatActivityWindowLabel, filterUniqueContributors). The 95-line routine that actually produces static/data/portal-contributors.json had no assertions.

What is covered

Driven through a stubbed globalThis.fetch, restored in a finally block:

• payload shape — generatedAt, activityWindowSince, activityWindowLabel, bluefinPulseUrl, and a custom windowDays
• repository selection — org listing plus the ublue-os/bluefin and ublue-os/bluefin-lts fallbacks, forks and nameless repos skipped, fallbacks de-duplicated against the org listing
• pagination — per_page=100 page walk, the & separator required because ?type=all is already present, short-page termination, non-array page termination, and the three-page cap on commit history
• the since= activity-window filter on every commit request
• contributor filtering — bots, authorless commits, empty logins, synthesised html_url, and de-duplication of one person across repos
• degradation — org listing failure still yields the fallback repos, a single failing repo does not discard the others (Promise.allSettled), and a Discussions outage leaves commit contributors intact
• discussion authors inside the window are added; stale and undated ones are not
• the twelve-person limit and the unavailable: true empty payload

Result

metric	before	after
line	32.38%	75.00%
branch	75.00%	95.74%
function	60.00%	89.47%

npm run test:coverage: 1017 passing, 0 failing, gate satisfied. npx eslint and npx prettier --check are clean on the changed file.

The remaining uncovered range is main (lines 171-226), whose 24-hour cache-freshness branch and fatal-fallback write cannot be exercised without an injectable output path — that is a production change and is deliberately out of scope, as noted in #1234.

Disjointness

No other open PR touches scripts/fetch-portal-contributors.js or its test. Verified against the open documentation PRs: #1215 and #1233 (scripts/fetch-firehose.js), #1212 (scripts/fetch-github-sbom.js), #1213 (scripts/lib/card-feed-parser.mjs), #1226 (.github/workflows/test.yml + scripts/portal-*.test.js), #1227/#1228 (src/components/), #1216 (static/data/).

Related Issue

Closes #1234

---

Filed by quality agent (hold-gated mode). Human review required — do not merge without it.

— hive: agent=quality backend=copilot model=claude-opus-5 copilot=1.0.78