Repository navigation
ci(website): deploy on nostream-website release publish, not on main pushes - #834
chappie-daemon wants to merge 6 commits into
Conversation
…pushes The site now deploys when the changesets workflow tags and publishes nostream-website@X.Y.Z (the release event this workflow filters on). Merging relay work to main no longer touches the site deploy; a site deploy happens exactly when a release PR that bumps nostream-website merges. Checkout on a release event is the tagged commit, so the deployed assets are the released build. Relay image releases keep their v* gate in release.yml — the tag namespaces do not overlap. workflow_dispatch remains for manual deploys of main.
🦋 Changeset detectedLatest commit: 40540a4 The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
[High impact] The changes appear safe to merge; the queue change addresses the previous finding without weakening deployment ordering.
|
Greptile P1: with the release trigger, a v* relay release entered the workflow-level deploy-website concurrency queue before the job's tag filter rejected it, and as a queued entry it could replace a pending website deploy (cancel-in-progress only protects the running one). Concurrency moves to the job so rejected events never queue.
…eploy # Conflicts: # .github/workflows/deploy-website.yml
|
Superseded by #837, which carries the remaining fixes this branch had against main: the checkout pinned to |
Change
Deploy Website no longer triggers on pushes to main. It now triggers on release published, filtered to
nostream-website@tags.Flow after this change:
nostream-websitechangeset → PR → merge to main.changeset-release/main).nostream-website@X.Y.Zand publishes the GitHub release.Relay-only merges (e.g. #830) never deploy the site. Relay image releases keep their
v*gate in release.yml; the tag namespaces (v*vsnostream-website@*) don't overlap.workflow_dispatchstays for manual deploys of a chosen ref.Note on the currently-failing run (38797428680)
That failure is independent of this change:
check-hero-nipsfails because #830 added NIP-77 (supportedNips24 → 25) and the hero film still shows 24. The film re-render runs on the framefields GPU pipeline — the gpu-render container + authz carve-out from 2026-10-10 are no longer present on the host, so the render is currently blocked from the sandbox and needs the host-sidecontainers/gpu-render/gpu-render-setup.sh --applyre-applied (or another GPU host). Until the film is re-rendered, the next nostream-website release will fail its site build at the guard.