vp run forwards no signal to its task, so stopping vp alone orphans the task #3090

Description

vp run <script> forwards no signal to the task it runs. A SIGINT or SIGTERM sent to the vp process alone ends vp, and the script's process keeps running, reparented to init or the user manager. A Ctrl-C in a terminal works only because the terminal sends it to the whole process group. Tools that stop a job by signalling the process they started leave the task running. That includes kill <pid> from a script, a supervisor, or an editor or agent harness that started vp run dev as a background command.

Tested on vite-plus 1.2.0 (global vp and the local CLI), on Linux:

This looks deliberate. vite-task at 9ba4cc93, crates/vt/src/session/mod.rs:382-388, says "Child tasks receive the signal directly from the terminal driver". The docs don't mention it, though.

Suggested solution

Forward SIGINT, SIGTERM and SIGHUP from vp to each running task, then wait for the tasks to exit. This is what npm run does (@npmcli/run-script lib/signal-manager.js), and what Turborepo (turborepo-signals) and concurrently (kill-on-signal.ts) do.

The global vp would also need to pass the signal on to the local CLI it starts.

Alternative

Document that vp run must be stopped by signalling its process group, so that harnesses and scripts can do that.

Additional context

We hit this with a long-running dev script that supervises several servers. Stopping vp alone left all of them running and holding their ports. We now tell agents to signal the script's own process instead.

Validations

• Read the Contributing Guidelines.
• Confirm this request is for Vite+ itself and not for Vite, Vitest, tsdown, Rolldown, or Oxc.
• Check that there isn't already an issue requesting the same feature. I searched vite-plus and vite-task for signal, SIGINT, SIGTERM, kill, orphan and process group.