feat(daemon): integration tests and bug fixes for recovery paths

[zhubert]

Summary

• Add daemon integration test suite exercising the full tick() loop with controlled mocks (16 tests covering happy path, failure modes, and recovery)
• Add reusable FakeProvider test double implementing all 9 provider interfaces
• Fix 3 bugs found during integration testing

Bugs Fixed

Terminal state step name lost — ProcessStep returned empty NewStep for succeed/fail states, causing CurrentStep to be cleared to "". The dashboard and erg status now correctly show "done"/"failed" instead of blank.

Docker recovery spawns duplicate workers — processIdleSyncItems didn't filter async actions (ai.code, ai.fix_ci, etc.), so after Docker recovery set phase=idle on a coding item, it re-executed the async action via executeSyncChain, creating a second session and worker for the same issue.

docker_pending items stuck forever — resumeDockerPendingItems only ran on the dockerDown→healthy transition. Transient Docker blips (worker fails but next health check passes) left items permanently stuck in docker_pending.

Test plan

• go test -p=1 -count=1 ./... — all packages pass
• goimports -w . / go vet ./... / staticcheck ./... — clean
• 16 integration tests cover: happy path (issue→merge), worker failure, max concurrency, dedup, external close, Docker error/recovery/down/transient blip, config save pause/recovery, retry delay elapsed/pending, idle sync recovery, feedback worker failure, duplicate worker prevention

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR adds a daemon integration test suite that exercises the full tick() loop end-to-end with mocked external dependencies, introduces a reusable FakeProvider test double for issue providers, and includes fixes to recovery-path bugs uncovered by the new tests.

Changes:

• Add integration tests covering happy path, failure modes, and recovery behaviors across the daemon tick loop.
• Introduce internal/issues/FakeProvider implementing provider interfaces for deterministic integration testing.
• Fix workflow/daemon recovery behavior: preserve terminal step name, avoid duplicate async work after Docker recovery, and resume docker_pending items when Docker is healthy.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
internal/workflow/engine.go	Preserve NewStep for terminal states so daemon/dashboard don’t lose the terminal step name.
internal/workflow/engine_test.go	Add regression test ensuring terminal states keep their step name.
internal/issues/fake_provider.go	Add a configurable fake provider implementing multiple provider interfaces for tests.
internal/daemon/processing.go	Harden recovery paths: skip re-running async actions in idle sync recovery; resume docker_pending items whenever Docker is healthy.
internal/daemon/integration_test.go	Add a full tick() integration test suite with mocked git/gh behavior and worker completion simulation.

Comments suppressed due to low confidence (1)

internal/workflow/engine.go:99

• ProcessStep() now preserves NewStep for succeed/fail states, but NewPhase is still left empty. Daemon.executeSyncChain/handleAsyncComplete call DaemonState.AdvanceWorkItem(result.NewStep, result.NewPhase), so an empty NewPhase will overwrite the persisted WorkItem.Phase with "" when a terminal state is processed. Consider setting NewPhase to item.Phase (or "idle") for terminal states so phase doesn’t get cleared as a side effect of reaching done/failed.

	switch state.Type {
	case StateTypeSucceed:
		return &StepResult{
			NewStep:    item.CurrentStep,
			Terminal:   true,
			TerminalOK: true,
			Hooks:      state.After,
		}, nil

	case StateTypeFail:
		return &StepResult{
			NewStep:    item.CurrentStep,
			Terminal:   true,
			TerminalOK: false,
			Hooks:      state.After,
		}, nil

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.