Skip to content

fix: harden fork maintenance and use manual candidate builds - #3

Merged
LingzheZhao merged 1 commit into
mainfrom
codex/maintenance-squash
Sep 20, 2026
Merged

LingzheZhao merged 1 commit into
mainfrom
codex/maintenance-squash

Conversation

@LingzheZhao

Copy link
Copy Markdown
Member

Replace the original merge of #2 with a GitHub squash merge, as requested by the project owner. This retains task-control authorization and archive-extraction fixes, standalone candidate packaging, and local-first checks. GitHub Actions retains only the manually dispatched distribution build.

The complete source tree is identical to the original maintenance head 4a0d27bcd79f1d16a25342495a724a87b847d130. Its workstation acceptance passed: exact-source Linux builds and wheel, real static CPU task, core task-control authorization regression, 1 quick regression, and 27 archive tests. Full evidence and artifact-reuse limits remain recorded in #2. This history rewrite changes no product files and does not require repeating those tests.

The original merge commits are being replaced in maintenance-then-pools order. The previous main tip is preserved locally at codex/backup-main-before-squash-20260920.

Reapply the maintenance changes from PR #2 for GitHub squash integration.
The source tree matches the workstation-validated 4a0d27b revision.
@LingzheZhao
LingzheZhao merged commit cef5936 into main Sep 20, 2026
2 of 7 checks passed
LingzheZhao added a commit that referenced this pull request Sep 28, 2026
Trials and Generic Tasks can now open a task/allocation resource dashboard backed by the existing cluster-setup Prometheus/Grafana stack. The integration is opt-in via `integrations.grafana_task_resources` and disabled by default.

A small authenticated UI configuration route returns only the validated dashboard URL and cluster selector. Trial details and the shared task logs page pass the real task ID, optional single allocation, and task start/end timestamps to dashboard UID `det-task-resources`. All allocations use Grafana's `$__all`; running tasks use `to=now`. URL encoding preserves Grafana subpaths, and unsafe schemes/embedded credentials are rejected. Cookie-only sessions work, unavailable optional configuration leaves task management intact, and navigation clears stale task metadata.

Generic Tasks currently have no dedicated details/Job Queue page; the existing `/det/generic/<task-id>/logs` route provides the entry. Grafana permissions remain independent; this change does not introduce a resource query API or claim per-task isolation in Grafana.

Validation: Go config tests (including recursive integration validation), master/internal compilation, Node 20.20.2 focused URL tests (3), full TypeScript check, focused ESLint/Prettier/Stylelint and diff checks passed. The companion provisioned dashboard was exercised through real Prometheus/Grafana with synthetic data and visually checked in a browser. No Actions run was triggered. Actual lab exporter access and production enablement remain pending; the integration stays off until configured.

Companion: [cluster-setup PR #3](WU-CVGL/cluster-setup#3) supplies collection fixes, identity rules, task dashboard and scrape credential migration. One coherent PR per repository.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant