[EricTechPro]

Note

⏳ In review · round 1 · ✅ 6/7 AC · Closes #64 · Closes #65 · head 0511f63

Problem

• 🐛 [bug] collect: the filer cannot run on Windows (fcntl, exec of a .sh, fd lock token) #64 Where: scripts/super-board-collect-state.py, run by super-collect-file.sh on Windows 11 + Git for Windows + native Python
  • a. Run super-collect-file.sh --config <cfg> --type bug --source prs ... --yes
  • b. The helper dies on import fcntl (ModuleNotFoundError)
  • c. Past that, os.execvpe of the .sh filer fails (Exec format error), and the child's lock check needs an inherited fd that MSYS bash does not pass on
• review: refactor tickets filed by the Reviewer get their title prefix twice #65 Where: scripts/super-review-file-refactor.sh, called by the Reviewer
  • a. The Reviewer passes --title "♻️ [refactor] images: one image-with-fallback component"
  • b. The filer prefixes again: ♻️ [refactor] images: ♻️ [refactor] images: ...
• Who: Windows installs of super-collect (🐛 [bug] collect: the filer cannot run on Windows (fcntl, exec of a .sh, fd lock token) #64); every board that gets Reviewer refactor tickets (review: refactor tickets filed by the Reviewer get their title prefix twice #65)

Solution

• collect-state imports fcntl only when present; otherwise it locks with msvcrt.locking and turns contention into BlockingIOError, so the existing retry and exit-4 path is unchanged
• Without fcntl, the helper runs the .sh filer as a bash child and keeps holding the lock until it exits, instead of exec
• Without fcntl, the child proves the lock with a per-hold random token file (removed when the hold ends) instead of an inherited fd; POSIX keeps flock, exec and the inode check unchanged
• scheduled-intake.md names both lock paths; the module docstring states the platform behavior
• The refactor filer strips a leading <emoji> [<kind>] <scope>: from --title before adding its own; only an all-non-ASCII first word triggers it
• super-review/SKILL.md says --title is the bare problem; the stale "lands in Backlog, not Ready" header comment now matches the code

Acceptance criteria

• The collector helper imports and runs with no fcntl module
  ↳ tests/test_collect_windows.py ImportTests + every intake case through a no-fcntl shim
• A concurrent collector call returns collector busy; no writes attempted (exit 4) on the portable path
  ↳ WindowsIntakeTests.test_bounded_lock_wait_does_not_mutate_and_can_run_after_release
• An interrupted write still blocks retry with "Collector write outcome unresolved"
  ↳ WindowsIntakeTests.test_interrupted_write_still_blocks_retry
• Receipt actions refuse a missing or forged lock token
  ↳ WindowsIntakeTests.test_receipt_action_refuses_forged_or_stale_token
• Dry-run and a real --yes filing exit 0 on real Windows 11 + Git for Windows
  ↳ not verified: no Windows host here; the msvcrt path is exercised through a flock-backed stand-in
• A pre-formatted refactor title is not prefixed twice
  ↳ tests/test-file-refactor.sh "a pre-formatted title is not prefixed twice (review: refactor tickets filed by the Reviewer get their title prefix twice #65)"
• Bare titles such as Map [x] fields: keep order are left untouched
  ↳ same block, two bare cases

Iteration history

Lane	Done	Time	Details
🔨 builder	✅	Oct 18	bash tests/run-safety.sh → PASS 58 suites · 0511f63

Risk

🟢 Low · POSIX lock path makes the same flock, exec and inode calls as before, behind if fcntl; the Windows path is new and tested only through a stand-in. The title strip only fires on an emoji-led [kind] scope: prefix.

🤖 Generated with Claude Code

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →