[quality] test: unit coverage for markdown-generator report-section builders (generateBuildHealthSection, generateBotDetailsList, generateCategorySectionWithSubsections)

[kubestellar-hive]

Test Improvement

Adds scripts/markdown-generator-sections.test.js — 24 cases, no production code changed.

Exact cluster claimed: the three exported report-section builders in scripts/lib/markdown-generator.mjs that scripts/report-markdown.test.js never imports — generateBuildHealthSection, generateBotDetailsList, and generateCategorySectionWithSubsections (which reaches the private filterItemsByLabels and formatItemList). No other file is touched.

What is covered

generateBuildHealthSection

• the null / undefined / missing-images / empty-images guards that return ""
• successes and failures derived from successRate × totalBuilds
• all four momChange branches: null → _Baseline_ (not a 0% badge), positive → success badge, 0 → the non-negative branch, negative → a critical badge with the shields.io double-dash escape (--7%25, not -7%25)
• the 100% Club member list and its _None_ fallback
• the images.find(...)?.totalBuilds || 0 fallback when stats.mostActive names an image no longer in images
• Math.round(avgDuration / 60) minute rendering

generateBotDetailsList

• flattening items across several bot-activity entries
• MDX curly-brace escaping, so a Renovate title like bump {pkg} to {version} is not read as JSX
• the missing-author unknown fallback
• the zero-width space in [@\u200B<author>] that keeps the published report from notifying the author

generateCategorySectionWithSubsections

• the whole-section > Status: _ChillOps_ collapse and the per-subsection ChillOps markers
• label matching under an emoji-prefixed category name, and rejection of an item that neither labels nor smart categorization claim
• displayedUrls recording, and the de-duplication path that drops an already-displayed item
• the ##### Dakota (GNOME OS Prototype) grouping in formatItemList, including ordering after the non-Dakota items and its absence when no Dakota item exists

Coverage delta

scripts/lib/markdown-generator.mjs:

	before	after
line	71.89%	81.51%
branch	48.08%	65.22%
functions	69.70%	87.88%

Full suite on a clean checkout: 949/949 pass (was 925). npx eslint and npx prettier --check are clean on the new file.

Related Issue

Refs #1168 — that issue also asks for scripts/lib/sbom/api.js (claimed by #1210) and for an injectable-client seam in graphql-queries.mjs / tap-promotions.mjs, which is a production change outside what this agent may PR. Its first three targets are in fact already covered; see the comment on the issue.

Disjointness

Checked against every open documentation PR: #1210 (scripts/lib/sbom/api.test.js), #1213 (scripts/lib/card-feed-parser.mjs), #1215 (scripts/fetch-firehose.js), #1226 (.github/workflows/test.yml + portal tests), #1227 (src/components/analytics/). None touches scripts/lib/markdown-generator.mjs or scripts/markdown-generator-sections.test.js.

---

Filed by quality agent (hold-gated mode). Human review required.

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

[hanthor]

Verified these tests are not vacuous — they hold real coverage.

Test-only PRs are the easy place to hide assertions that pass no matter what the source does, so I checked rather than took it on faith.

Coverage genuinely moves

main:    markdown-generator.mjs | 71.89 | 48.08 | 69.70
+pr1229: markdown-generator.mjs | 81.51 | 65.22 | 87.88
                                  lines   branch  funcs

Functions from 69.70% → 87.88% is the number that matters for a PR claiming to cover three previously-unimported builders — it confirms they really were unreached before.

Full suite merged into current main:

$ npm run test:coverage   # origin/main + pr1229
ℹ tests 949
ℹ pass 949
ℹ fail 0

Baseline on unmodified main is 925 / 925 / 0, so 24 net new tests, all passing, no regression.

Mutation-tested: both mutants killed

I broke the source in two places and confirmed the new tests catch it.

Inverting the ChillOps guard in generateCategorySectionWithSubsections:

$ sed -i '720s/planned.length === 0 && opportunistic.length === 0/false/' scripts/lib/markdown-generator.mjs
$ node --test scripts/markdown-generator-sections.test.js
(exit 1)

Zeroing the successes arithmetic in generateBuildHealthSection:

$ sed -i '1200s|Math.round((img.successRate / 100) * img.totalBuilds)|0|' scripts/lib/markdown-generator.mjs
$ node --test scripts/markdown-generator-sections.test.js
(exit 1)

Restored source, clean run exits 0. So the tests are pinned to actual behaviour — the assert.match(md, /\| \bluefin` | 75% | 30 | 10 |/)` style of asserting on rendered output rather than on the fact that a string was returned is what makes that work.

No findings. The header comment explaining why these builders matter ("these builders render the published monthly report, so a regression shows up on the site rather than in CI") is a good habit.

---

Generated by Claude Code