Validate and update links (STF-557)

[oschwald]

Validated every link in the repository with lychee (max_redirects = 0, so any redirect surfaces as an error) and updated those that were out of date or redirecting.

CI setup

• Added lychee.toml tuned for this C library (Markdown docs + C sources/headers; excludes vendored submodules, the t/ test harness, build/generated dirs, the Hugo site output, and the Changes.md changelog).
• Added a Links GitHub Actions workflow (.github/workflows/links.yml) that runs on push, pull request, weekly schedule, and manual dispatch; opens an issue on scheduled failures.
• Added .lycheecache to .gitignore.

Link fixes

README.dev.md

• https://help.launchpad.net/Packaging/PPA -> https://documentation.ubuntu.com/launchpad/user/Packaging/PPA/
• https://github.com/Homebrew/homebrew-core/edit/master/Formula/libmaxminddb.rb -> https://github.com/Homebrew/homebrew-core/blob/main/Formula/lib/libmaxminddb.rb (formula relocated under Formula/lib/, default branch renamed master -> main; the original /edit/ URL also forced a login redirect, so the blob page is used and the surrounding "Edit the file" instruction still applies)

README.md

• https://ports.macports.org/port/libmaxminddb -> https://ports.macports.org/port/libmaxminddb/ (trailing slash)

No changelog entries were modified.

Final lychee result

26 Total, 22 OK, 0 Errors, 4 Excluded

Part of STF-557.

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request configures the Lychee link checker by adding a lychee.toml configuration file, ignoring .lycheecache in .gitignore, and updating several outdated URLs in the documentation. The review feedback suggests anchoring the regular expressions in exclude_path to prevent over-broad exclusions (such as 't/' matching any path containing t/) and removing HTTP 5xx status codes from the accept list to avoid silently ignoring permanently broken links.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

The regular expressions in exclude_path are unanchored, which can lead to accidental exclusions of valid files.

Specifically, 't/' matches any path containing the substring t/. This will exclude any directory or file path that ends with t followed by a slash, such as docs/content/, docs/layouts/, src/test/, or any path containing words like client/, asset/, project/, net/, etc. This silently disables link checking for a large portion of the repository.

Similarly, 'build' will match any path containing the word build (e.g., dev-bin/build-release.sh or any other file with build in its name).

To prevent these over-broad matches, anchor the regular expressions to the start of the path (^) and ensure directory boundaries are respected.

exclude_path = [
    '^\.git/',
    # Build / generated directories
    '^build/',
    '^\.libs/',
    '^autom4te\.cache/',
    '^generated/',
    # Generated Hugo site output
    '^docs/public/',
    # Vendored submodules and test harness/fixtures
    '^maxmind-db/',
    '^t/',
    # Changelog: historical entries are preserved as-is, not rewritten
    '^Changes\.md$',
]

[oschwald]

Fixed — the exclude_path patterns are now segment-anchored as (^|/)dir/, so short names like t no longer match unintended paths.

— Claude (posted on Greg's behalf)

[gemini-code-assist]

Including "500..=599" in the accept list means that any HTTP 5xx server errors (such as 500 Internal Server Error, 502 Bad Gateway, or 503 Service Unavailable) will be treated as successful checks.

While this prevents temporary server issues from failing the CI build, it also silently ignores permanently broken links where the hosting server is misconfigured or down.

Instead of accepting 5xx status codes globally, it is recommended to rely on Lychee's retry mechanism or exclude specific flaky domains if necessary.

Suggested change

	accept = ["100..=103", "200..=299", "403", "429", "500..=599", "999"]
	accept = ["100..=103", "200..=299", "403", "429", "999"]

[oschwald]

Keeping 500..=599 in accept, matching the dev-site and blog-site configs — transient upstream 5xx shouldn't fail link-checking CI.

— Claude (posted on Greg's behalf)

[horgh]

[minor] Launchpad link points at a client-side redirect stub, not the canonical page Line 23: new URL https://documentation.ubuntu.com/launchpad/user/Packaging/PPA/ returns HTTP 200 with no Location header (so lychee max_redirects=0 passes), but the page body is a Read-the-Docs redirect stub: meta-refresh + window.location.href to a relative target that ultimately 302->200 at https://documentation.ubuntu.com/launchpad/user/reference/packaging/ppas/ppa/ ('Personal Package Archive - Launchpad manual', real PPA content). It works for a JS-enabled browser, but it's a hidden client-side redirect — exactly what this PR's max_redirects=0 policy is meant to surface and replace with a canonical URL. Recommend linking the canonical page directly: https://documentation.ubuntu.com/launchpad/user/reference/packaging/ppas/ppa/ (verified curl: 200, no redirect).

[oschwald]

Good catch — fixed in 29c3d32. Linked the canonical page directly (…/launchpad/user/reference/packaging/ppas/ppa/, verified 200 with no HTTP or client-side redirect). You're right that max_redirects = 0 only catches HTTP redirects, not the JS/meta-refresh stub.

— Claude (posted on Greg's behalf)