rail-preview

Staging copy of the lab website's wvu branch and the DSPIRA lessons site's wvu branch (the WVU Design System builds), published together at https://wvurail.org/rail-preview/ and https://wvurail.org/rail-preview/dspira-lessons/ for the Strategic Communications review. Not the real site, and not indexed: every page carries noindex.

There is no content here. The one workflow in .github/workflows/ checks out WVURAIL/wvurail.github.io@wvu and WVURAIL/dspira-lessons@wvu, builds them under /rail-preview/ and /rail-preview/dspira-lessons/ with the same gems GitHub Pages uses, and deploys the result. It runs every half hour and on demand (Actions → Stage the wvu branch → Run workflow).

When the branch has been merged and the site is live on rail.wvu.edu, delete this repository.

Preview validation

Pull requests build and validate the combined preview without deploying it. The link check runs after both sites are assembled and the lab-to-lessons links are rewritten:

npm ci
npm test
python3 -m pip install -r requirements-checks.txt
python3 -m unittest discover -s tests -v
python3 scripts/combine_sitemaps.py --site _site
node scripts/trim_styles.mjs _site
python3 scripts/check_links.py --site _site
python3 scripts/check_pages.py --site _site
python3 scripts/check_readability.py --site _site

The checker verifies local link and asset targets under /rail-preview/, including relative links, redirects, and extensionless GitHub Pages URLs. Other sites on the host, such as /lightwork/, stay at their published paths and are outside this offline check. External URLs and heading fragments are covered by the source repositories' link audits.

The page check validates language, preview noindex, unique titles and descriptions, title and description lengths, Open Graph tags, structured data, heading order, image alt attributes, and focusable skip-link destinations. Redirect stubs retain the language and noindex checks but do not need content-page metadata.

The readability check limits prose sentences to 20 words. Published titles, citations, code, and quotations are preserved. Exact text and page exceptions are recorded in scripts/readability_exceptions.json; changed passages must be reviewed again. Equations and inline code count as single tokens.

The combined sitemap at /rail-preview/sitemap.xml includes both sites while preserving their exclusions. Use this URL when configuring the preview crawl. Siteimprove SEO discovery also depends on the host's root robots.txt, which currently advertises only the production sitemap. See the audit notes for this preview limitation.

Use Node.js 22 for the stylesheet step. It trims the published Design System CSS against the built HTML and JavaScript, keeping menu states, fonts, animation definitions, and CSS variables. The source bundles remain unchanged, and each build recomputes the styles needed by the current pages.

See the September audit notes for the source repositories, report dispositions, and remaining review items. Merge the source-site changes before enabling these stricter checks against their remote wvu branches.