linkchecker doesn't detect broken links on the same domain that miss the library prefix
When checking http://domain/library/* then all links to http://domain/somethingelse are seen as external and ignored. However, this causes a specific error case not to be seen: url_for links that have a non-existing destination.
For example: url_for('/install/valid.html') ==> http://domain/library/install/valid.html url_for('/install/invalid.html') ==> http://domain/install/invalid.html <-- not reported
Linkchecker offers --check-extern (to proceed to "external" links), and with either --ignore-url REGEX or --no-follow-url REGEX we should be able to keep it to the same domain.
It's probably a good idea to start checking external links (without following) as well.
For now, a simple Nginx workaround solves the immediate issue (by allowing use of linkchecker http://localhost/index.html):
location /index.html { return 302 sub/index.html; }
location / { deny all; return 404; }