Exit codes

check and build are meant to be run by a pipeline, so what they report matters less than what they return.

The codes

Errors stop the build

An error means a reader would hit something that is not there — a link to a page that was never published, an image that does not exist, a wikilink matching nothing. When any is present, build writes no output at all.

That is a deliberate contract rather than an implementation detail: a site published with half its images missing is worse than a site that did not publish, because nobody finds out until a reader does.

See Error messages for what each one says and how to read it.

Warnings do not

A warning is something that depends on context the checker does not always have. A root-absolute reference like /assets/logo.png is right if the site is mounted at the root of a domain and wrong if it is served from a sub-path. Ordinarily the checker cannot tell which — but when siteUrl already declares a sub-path mount, it warns about a root-absolute reference even if the reference resolves today, since that is the one case it can actually judge. It reports and continues either way.

In a pipeline

- run: npx canopy-page check docs/site
- run: npx canopy-page build docs/site -o dist/help

The first step is the fast one: it never renders. Putting it before an expensive job means a broken link costs seconds rather than a full build.