GitHub Pages
GitHub Pages (Ctrl+0) shows the site a
repository publishes, if it has one: one row per publish, with whether
it worked, the commit it came from, who pushed it, and how long it
took.
The status bar leads with what you most likely came to check, for example “Live at https://owner.github.io/name/ — built, published from main/docs.” The details panel adds the rest of the site’s settings: who builds it, whether HTTPS is enforced, any custom domain, whether it is public, and whether it has its own 404 page.
If the repository has no site, the details panel says “GitHub Pages is not enabled for this repository.” rather than leaving you with an empty list, which would look the same as a site that has never published.
The pages a site serves
Enter in the publish history lists Published
Pages: every file the site serves, each with the address it is
served at.
| Key | Action |
|---|---|
Enter |
Open the page in your browser |
S |
Open the site’s home page in your browser (also in the publish history) |
F |
Add the page to favorites, so one you check often is a keystroke away |
Backspace |
Return to the publish history |
Actions → Open Published Site does what
S does, once GHManage knows there is a site.
Two things about this list are worth knowing, and the details panel points them out on the pages they apply to:
- Jekyll sites serve Markdown as HTML. A file stored
as
guide/setup.mdis served atguide/setup.html, and the address listed is the one that works. A site with a.nojekyllfile skips Jekyll and is served exactly as committed. - A site built by Actions is listed from its source, not its output. GitHub publishes those sites through a workflow, and the workflow decides what actually ships, which may be only some of these files, or files it generated that are not in the repository at all. For sites GitHub builds itself, the list is what is served.
The publish history has the same split. Sites GitHub builds itself keep a full build log with durations and error messages. Sites built by Actions have no build log, so their history comes from their deployments instead, which record whether each publish worked but not how long it took.