[entelecheia]

Part of #394.

What changes

dot sync resume rewrote a unit installed before #394 to --store, but neither service
manager reads a rewritten unit file until it is told to:

• macOS: launchctl load finds the job already loaded (a job that keeps failing is
  still loaded) and leaves launchd running the command line it cached. ResumeKind now
  runs launchctl unload <plist> first and ignores its error, the way InstallKind does.
• Linux: systemd keeps the ExecStart it read until daemon-reload. ResumeKind now
  runs systemctl --user daemon-reload before start <timer>.

So plain dot sync resume --store=<s> heals a migrated unit, and the pause && resume
recovery hint that dot sync status prints for a spawn-failed job works on both platforms.

The scheduler lifecycle test helper now pins resume as two commands (unload then load;
daemon-reload then start), so a resume that skips the reload fails the tests that record
service-manager argv.

Stale help

dot sync target said a local target is also recorded in the global user state for future
workspaces to inherit. Since #409 the target is written only to this workspace's store
(<workspace>/.dotfiles/sync/config.yaml); a workspace without one syncs to the detected
cloud folder. The sentence and docs/commands/dot_sync_target.md (make docs) say that now.

Verification

• dot admit -- env -u CLAUDE_CONFIG_DIR GOMAXPROCS=2 go test -p 2 -parallel 2 ./internal/syncer/... ./internal/cli/... passes.
• dot admit -- make lint: 0 issues.
• The tests bite: removing the unload step from ResumeKind fails
  TestSyncSchedulerLifecycle_InstalledKinds (every resume row), TestSyncPauseResumeReloadsSpawnFailedJobs
  and TestSyncResumeMigratesAUnitInstalledWithTheRetiredStoreFlag; restored, they pass.
• The Linux half is compiled but not run here (macOS): its argv expectation comes from the
  same shared helper and is exercised by the CI test run.

Release note: dot sync resume now reloads the scheduler unit, so a unit that still ran the removed --profile flag is healed by it on both platforms.

Round 2

CI's unit (ubuntu-latest) job caught a wrong expectation, not a wrong product: on Linux
ResumeKind returns the daemon-reload error before start runs, so the injected-failure
row could not end with --user start. schedulerLifecycleCommands now stops the expected
list at the failing command on platforms whose resume returns there (macOS keeps both because
the unload error is ignored and the failure lands on the load). Product code unchanged.

Verification for this round: GOOS=linux go vet ./internal/syncer/ and the native
go vet ./internal/syncer/ both pass. The focused go test row and make lint were
deferred three times by the host admission gate ("memory pressure warning", five-minute
recovery streak not completed), so they are unverified on this host this round; the ubuntu
unit job is the executable proof for the Linux expectation.

Round 3

CI's ubuntu unit job was still red at the same row, one assertion later:
TestSyncSchedulerLifecycle_InstalledKinds/resume_stops_after_push_failure now passed the
argv comparison and failed the non-vacuity guard, because schedulerLifecycleActionCount
did not recognise --user daemon-reload — which is the only command a failed Linux resume
records.

schedulerLifecycleActionCount counts that prefix now, and TestSchedulerLifecycleActionCount
pins the counter as a small table test that runs on every platform, so an unrecognised
command fails here instead of in the ubuntu job. Test-only; no product change.

Verification for this round: gofmt clean, native and GOOS=linux go vet ./internal/syncer/
pass. The focused go test run and make lint were deferred again by the host admission gate
(memory-pressure episode still open after ~40 minutes, both attempts refused with the same
Retry after: 23s), so they are unverified on this host and the ubuntu unit job carries the
proof of the Linux expectation.

Round 4

make lint (misspell, locale US) failed on one word: recognise -> recognize in the
counter's comment. That word only; no code, no test logic.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-09T19:29:54.947372Z	c312e71	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.