5.0 KiB
Lifecycle Optimization Implementation Plan
For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (
- [ ]) syntax for tracking.
Goal: Make CAD save, close, exit, and resource-release outcomes explicit and recoverable across normal, failed, cancelled, and forced-stop paths.
Architecture: Keep business completion and infrastructure cleanup as separate outcomes. Extend the existing resource journal and runtime result objects with phase evidence, give owned helper processes a bounded natural-exit window, and preserve cleanup failures instead of converting them into a successful-looking result.
Tech Stack: Python 3.12, PySide6, pywin32, pytest, existing BatchCadRuntime, IsolatedCadSession, supervisor protocol, and VBS worker evidence files.
Spec: Approved in chat on 2026-10-30 as Phase 1 of the project optimization.
Global Constraints
- Preserve the existing no-replay protection for uncertain project mutations.
- Never terminate an unowned process or clear evidence that cannot be verified.
- Keep normal successful replacement behavior and public service method signatures compatible.
- Every production change must have a regression test that fails before the change.
Task 1: Model and persist separate save/close evidence
Files:
- Modify:
src/dx_productivity/trial_task_center/batch_runtime.py - Modify:
src/dx_productivity/trial_task_center/isolated_cad.py - Modify:
src/dx_productivity/trial_task_center/batch_resources.py - Test:
tests/v1_1/test_batch_runtime.py - Test:
tests/v1_1/test_isolated_cad.py
Interfaces:
-
Consumes existing CAD shutdown records and project checkpoints.
-
Produces journal fields for save requested, save returned, readback verified, project closed, and cleanup outcome.
-
Step 1: Write failing tests for a close failure after a save request and for a cleanup record that reports business completion with infrastructure failure.
-
Step 2: Run the focused tests and verify they fail because the current record has no separate save evidence and the completed result hides cleanup failure.
-
Step 3: Add the smallest evidence fields with backward-compatible defaults and preserve the existing uncertain/recovery guard.
-
Step 4: Run the focused tests and verify both pass.
-
Step 5: Run the related supervisor and recovery tests before moving to the next task.
Task 2: Allow owned helper processes a bounded natural exit window
Files:
- Modify:
src/dx_productivity/trial_task_center/isolated_desktop.py - Test:
tests/v1_1/test_isolated_desktop.py - Test:
tests/v1_1/test_batch_runtime_windows.py
Interfaces:
-
Consumes owned process identities and existing Win32 job cleanup records.
-
Produces cleanup records that distinguish naturally exited descendants from forced termination.
-
Step 1: Write a failing test where the root CAD process has exited and an owned helper exits shortly afterward; assert no forced termination is requested.
-
Step 2: Run the test and verify the current implementation calls the force cleanup immediately.
-
Step 3: Add a bounded wait only for identities whose owned root has already exited; retain force cleanup after the deadline.
-
Step 4: Run the focused Windows tests and verify the cleanup record includes the wait and final state.
-
Step 5: Run the resource release test group.
Task 3: Separate diagnostic journal failures from the CAD Quit protocol
Files:
- Modify:
src/dx_productivity/trial_task_center/isolated_cad.py - Modify:
src/dx_productivity/trial_task_center/batch_release_log.py - Test:
tests/v1_1/test_isolated_cad.py - Test:
tests/v1_1/test_batch_runtime.py
Interfaces:
-
Consumes the existing owned CAD identity and project-close state.
-
Produces an explicit diagnostic-write warning while still attempting a valid Quit sequence.
-
Step 1: Write a failing test that injects a closing-record write failure and asserts Quit is still attempted when project state is safe.
-
Step 2: Run the test and verify Quit is currently skipped.
-
Step 3: Isolate diagnostic persistence errors from the normal Quit decision and record both outcomes.
-
Step 4: Run the focused tests and verify both normal and failed-record paths.
-
Step 5: Run all lifecycle and supervisor tests.
Task 4: Verify the phase and document field validation
Files:
-
Modify:
docs/testing/2026-10-26-application-lifecycle-investigation.md -
Create:
docs/testing/2026-10-30-phase1-lifecycle-optimization.md -
Step 1: Run the complete test suite and capture the count and warnings.
-
Step 2: Run
python -m compileall -q srcandgit diff --check. -
Step 3: Record the exact scenarios covered and the remaining real DxDesigner field checks.
-
Step 4: Review the final diff for accidental changes outside Phase 1.