public/soundy-bits
1
0
Fork 0
Code Issues Pull requests Projects Releases 5 Packages Wiki Activity Actions

fix(ci): forgejo-first release (gitlab can't block it) + headless builds #15

Merged
psk merged 1 commit from fix/release-mirror-sync into master 2026-06-29 19:47:52 +00:00
Conversation 0 Commits 1 Files changed 2 +121 -6
psk commented 2026-06-29 19:32:23 +00:00
Owner
Copy link

For 0.3.2.

Decouple: the release job did the Forgejo release AND the GitLab mirror in one job, so a slow/broken forgejo->gitlab push mirror (the <=60s tag-sync wait) failed the WHOLE release even though the Forgejo (primary) release already succeeded. Split the GitLab work (release + Pages) into a downstream mirror-gitlab job (needs: release) that fetches the artifacts via upload/download-artifact. A GitLab/mirror failure now shows RED on that job but is isolated -- it can't block/undo the Forgejo release. GitLab is NOT best-effort. Kept the active push-mirror kick + ~3min poll.

Headless builds: render-only x86_64 + aarch64, each archive ships a HEADLESS.md with the Vulkan-driver deps; added to the README download table.

Note: cross-job artifact-passing only gets its real test on the release run (fallback: pull assets from the Forgejo release if the runner lacks the artifact server).

For 0.3.2. **Decouple:** the release job did the Forgejo release AND the GitLab mirror in one job, so a slow/broken forgejo->gitlab push mirror (the <=60s tag-sync wait) failed the WHOLE release even though the Forgejo (primary) release already succeeded. Split the GitLab work (release + Pages) into a downstream `mirror-gitlab` job (needs: release) that fetches the artifacts via upload/download-artifact. A GitLab/mirror failure now shows RED on that job but is isolated -- it can't block/undo the Forgejo release. GitLab is NOT best-effort. Kept the active push-mirror kick + ~3min poll. **Headless builds:** render-only x86_64 + aarch64, each archive ships a HEADLESS.md with the Vulkan-driver deps; added to the README download table. Note: cross-job artifact-passing only gets its real test on the release run (fallback: pull assets from the Forgejo release if the runner lacks the artifact server).
fix(ci): forgejo-first release + headless render builds (gitlab can't block it)
All checks were successful
ci / build-and-test (pull_request) Successful in 8m56s
Details
fa4bfa69d7
 
The release job did the Forgejo release AND the GitLab mirror in one job,
so a slow/broken forgejo->gitlab push mirror (the <=60s tag-sync wait)
failed the WHOLE release even though the Forgejo (primary/"local") release
had already succeeded. Forgejo is the source of truth; GitLab is a
downstream mirror and must never block or undo it.

- Split the GitLab work (release + Pages) into a separate `mirror-gitlab`
  job (`needs: release`) that fetches the built artifacts via
  upload/download-artifact. A GitLab/mirror failure now shows red on THAT
  job but is isolated -- it can't block or undo the Forgejo release.
  GitLab is NOT best-effort: real failures still fail the mirror job.
- Keep the active push-mirror kick + ~3min poll so GitLab still publishes
  promptly when it can.
- Add render-only headless builds for x86_64 + aarch64
  (--no-default-features --features diagnostics), packaged with a
  HEADLESS.md that spells out the runtime deps (only a Vulkan driver), and
  listed in the README download table.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
psk merged commit 075d68f5b8 into master 2026-06-29 19:47:52 +00:00
psk deleted branch fix/release-mirror-sync 2026-06-29 19:47:59 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
public/soundy-bits!15
No description provided.