Terminating subprocess on Windows when "running" is flaky

[MaxmaxmaximusAWS]

"estrella": "^1.4.0",

This config:

estrella.build({
  watch: true,
  run: "node build/index.js" // this process not killing
})

Temporary solution:

let childProcess = null

estrella.build({
  onEnd: () => {
    if (childProcess) { 
      childProcess.kill('SIGINT')
    }

    childProcess = spawn('node', ['build/index.js'], {
      stdio: [process.stdin, process.stdout, process.stderr],
    })
  },
})

[rsms]

What OS are you using?
(Estrella uses Nodejs's child_process module and process groups on POSIX systems to allow for process tree control)

[rsms]

Also can you please provide a complete repro? (ideally as a gist or a zip or tar archive)

FTR, the signalling code is here:

estrella/src/exec.ts

Lines 272 to 320 in 7294055

	// signal sends sig to the underlying process and returns true if sending the signal worked.
	// mode defaults to "standard"
	//
	// If the signal is successfully sent (not neccessarily delivered) true is returned.
	// If the process is not running, false is returned (no effect.)
	// If the process has not been started, an exception is thrown.
	// If the signal is not supported by the platform, an exception is thrown.
	// If another error occur, like signalling permissions, false is returned.
	//
	signal(sig :Signal, mode? :SignalMode) :boolean {
	const p = this._checkproc()
	if (mode == "group") {
	// Signalling process groups via negative pid is supported on most POSIX systems.
	// This causes subprocesses that the command process may have started to also receive
	// the signal.
	try {
	process.kill(-p.pid, sig)
	return true
	} catch (_) {
	// will fail if the process is not in its own group or if its is already dead.
	// fall through to "proc" mode:
	}
	}
	return p.kill(sig)
	}
	// kill terminates the command by sending signal sig to the process and waiting for it to exit.
	// mode defaults to "group".
	//
	// If the process has not exited within timeout milliseconds, SIGKILL is sent.
	// The timeout should be reasonably large to allow well-behaved processed to run atexit code but
	// small enough so that an ill-behaved process is killed within a reasonable timeframe.
	// If timeout <= 0 then the returned promise will only resolve if and when the process exits,
	// which could be never if the process ignores sig.
	//
	async kill(sig :Signal="SIGTERM", timeout :number=500, mode? :SignalMode) :Promise<number> {
	const p = this._checkproc()
	if (!this.signal(sig, mode || "group")) {
	return p.exitCode || 0
	}
	if (timeout <= 0) {
	return this.promise
	}
	return this._waitTimeout(timeout, (_, resolve) => {
	log.debug(()=>`${this} kill timeout reached; sending SIGKILL`)
	p.kill("SIGKILL")
	return this.promise.then(resolve)
	})
	}

[MaxmaxmaximusAWS]

What OS are you using?
(Estrella uses Nodejs's child_process module and process groups on POSIX systems to allow for process tree control)

windows 10 of course

[rsms]

Ah, yeah you're going to have some issues if the process is not responding to hangup or int signals.
You could try WSL or run your build manually (either in a BAT script or from js in onEnd, but you'll likely run into similar issues as with estrella using nodejs.)

[MaxmaxmaximusAWS]

Why you not use this code in windows platforms? =) This fix works on me:

let childProcess = null

estrella.build({
  onEnd: () => {
    if (childProcess) { 
      childProcess.kill('SIGINT')
    }

    childProcess = spawn('node', ['build/index.js'], {
      stdio: [process.stdin, process.stdout, process.stderr],
    })
  },
})

i can use WSL or WSL2 or docker or my macbook, but we need fix the bug, not hide the bug)