Wait for a service while it makes progress, and for one already on its way

[donislawdev]

Package C of the stability review: waiting that does not abandon a service half way.

What changes for a person

• The step limit counts time without progress. --timeout and "Wait up to" on the plan sheet now give up on a step once the limit has passed since the service last reported progress, not since the step began. A restart of a service that takes longer than the limit to stop ends with the service running instead of stopped.
• An entry already on its way is waited for. A stop asked of a service already stopping sends nothing and waits. A start asked of a service still stopping waits for it to stop and then asks. An entry that never leaves the opposite pending state is never asked, and the step says so.
• A start the service does not survive fails at once, with the service's exit code in errorCode and its own code (for 1066) in the error text, instead of timing out after the whole limit.
• The second Ctrl+C takes effect while a step is being waited for. An interruption skips a forward step not asked yet. Steps that put things back keep waiting.
• Two new plan warnings: starting a disabled entry (disabledCannotStart) and starting a paused one (pausedCannotStart).
• The note about the manager taking longer than --timeout to answer is read off the time of the answer alone.
• The window's reason for offering Force stop no longer names a number of seconds that could be wrong, and the line under a waiting step says the limit is about progress.

Contract

• The meaning of --timeout (README, site and docs mirrors updated).
• Two warning kinds, added.
• errorCode on a failed start carries the service's exit code. No new fields. An abandoned step reports timedOut.

Checked

• Core, command line, window (plan sheet, sentences, text guards), architecture, site and the integration tests that run anywhere - the classes this touches, all green.
• 41 mutation entries proved: 19 new, 6 re-anchored, the rest the neighbours in the runner. One neighbour was missed at first and its test was made to reach the fault.
• On this machine, reading only: the exit codes read through the service manager matched sc query on 800 of 800 entries, and bws start on a disabled service shows the new warning. A restart of Spooler ran through the new path with both steps succeeding.

Not checked

• A real service that takes longer than the limit to stop, and a real service failing at start - the throwaway machine did not answer. Both are covered on the fake manager only, including the manager refusing a request to an entry still on its way.
• The window on a live run.

🤖 Generated with Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Repository UI (base), Organization UI (inherited)

Review profile: ASSERTIVE

Plan: Advanced

Run ID: e90457f5-d8ba-499b-97a5-5dd9cd471af6

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}