Skip to content

[ADHD] v0.6.0 — scheduled Git fetch and post-push refresh #52

Closed

Description

@uniskela

Parent: #51

Goal

Remove routine manual repository fetching while preserving every draft and unpushed Git commit.

Focus

Verify per-repository automatic fetch settings on current main: opt-out and a custom interval in Overview → Automatic fetching, then close #52.

Next

Resume

Check out main at or after da55d99 and open Overview → Automatic fetching for one repository. docs/internal/automatic-fetch-review.md has the timing, migration and queue/lock notes.

Done

Remaining completion gate

No release, image publication or production deployment has been performed. #52 remains open.

Made with ADHD Progress Hub

Activity

  1. uniskela commented on Oct 20, 2026

    @uniskela
    OwnerAuthor

    Done

    PR #59 remediation is pushed at 1848cb5: deterministic post-push race fix with one durable coalesced job, workspace/repository failure isolation, dedicated attempt/failure/change timestamps and focused repository status. Local check/build/migration/audit passed; Playwright 33 passed with 9 intentional platform skips. Controlled Docker integration: 61 passed; public-clone smoke, restart, disabled worker and missing-key rejection passed. All original Kody threads have individual evidence-based replies; another code/business review is requested.

    Next

    Review new CI/Kody results on #59. Implement repository-specific fetch policy in immediate follow-up #60; this issue remains open until that acceptance work is done. Security findings are being handled separately, without mixing changes into #59.

    Resume

    Open PR #59 at 1848cb5 and its validation comment; read docs/internal/automatic-fetch-review.md for queue semantics, benchmark decision and crash-lock limits. Continue from the first outstanding CI/review item, then #60. Do not merge/release or deploy production without approval.

  2. uniskela commented on Oct 20, 2026

    @uniskela
    OwnerAuthor

    Done: PR #59 remediation is pushed at bc85555; all current GitHub checks, Kody review and Sonar quality gate pass. All three original Kody threads are answered/resolved; business validation found no MUST_FIX findings. Dedicated sync timestamps, durable post-push follow-up and failure isolation are implemented. Verification: 377 tests; browser 33 passed/9 intentional skips; controlled rootless-Docker integration 61 passed, plus build/migrations/smoke checks.

    Next: review #59. Per-repository policy remains the immediate follow-up #60; #52 stays open for that work. Independent security hardening is #61 (8266e33), also all-green.

    Resume: start with docs/internal/automatic-fetch-review.md on #59, then #60's acceptance criteria. No merge, release or production deployment was performed.

  3. uniskela commented on Oct 20, 2026

    @uniskela
    OwnerAuthor

    Drift check: PR #62 merged on 2026-10-20 (da55d99, closing #60) after this issue was last edited, so Focus/Next/Resume now point at verifying the per-repository fetch settings on main before closing #52. Both fetch PRs are in pending release PR #41 (v0.6.0).

  4. uniskela commented on Oct 22, 2026

    @uniskela
    OwnerAuthor

    Verified on main (8333dd8), release audit #57

    Acceptance criteria checked against the code and tests on main:

    • Scheduled fetch: respects STACK_MANAGER_WORKER_ENABLED and STACK_MANAGER_AUTO_FETCH_ENABLED, and waits a full interval after a failure. Duplicate jobs are prevented by the active-job dedupe key, and overlapping scans are coalesced. Unit tests pass, plus a new scanner test (fix: harden v0.6.0 release audit labels, tests and upgrade docs #67) proving nothing is scanned or queued when the worker or server auto-fetch is off.
    • Per-repository policy: enable/disable, nullable interval that inherits the server default, validation 5–1440. Playwright desktop/mobile passed. The server-disabled scenario was run separately with STACK_MANAGER_AUTO_FETCH_ENABLED=false and passed.
    • Post-push refresh: a durable coalesced job (follow-up if a fetch is running); a queue failure never changes the push result. Integration tests pass.
    • Safety: fetch only updates refs/remotes/origin/*. Drafts (DB) and unpushed commits (refs/stack-manager/heads/*) survive failed fetches and the v0.5.0 → v0.6.0 upgrade (verified with a database seeded by v0.5.0 code). A new integration test shows one repository's failing fetch does not block another.
    • Restart: pending sync jobs left by v0.5.0 are executed by the v0.6.0 worker after upgrade.

    Evidence: pnpm check 656/656 on main, 667/667 with #67; full browser/axe 43 passed / 13 intentional skips; report docs/internal/releases/v0.6.0-readiness.md in #67.

    Not verified here: a live external push picked up by polling on the operator's own instance. That is step 4 of the v0.6.0 acceptance checklist, tracked in #57.

    Closing as verified on main.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions