Concurrently crashes leaving processes running when piping command exits

[idrise]

Recreate

concurrently --kill-others -n api,ui \"nx serve api\" \"nx serve ui\" | lnav -t

Then press q to quit lnav and this error is reported and both API and UI continue to be served in the background. It would be great if concurrently shut down when the write stream is destroyed.

    setTimeout(function () { throw err; }, 0);
                             ^

Error [ERR_STREAM_DESTROYED]: Cannot call write after a stream was destroyed
    at Socket.Writable.write (_stream_writable.js:322:17)
    at Logger.log (***/node_modules/concurrently/src/logger.js:107:27)
    at Logger.logCommandText (***/node_modules/concurrently/src/logger.js:74:21)
    at Logger.logCommandEvent (***/node_modules/concurrently/src/logger.js:69:14)
    at SafeSubscriber._next (***/node_modules/concurrently/src/flow-control/log-exit.js:8:25)
    at SafeSubscriber.__tryOrUnsub (***/node_modules/rxjs/internal/Subscriber.js:205:16)
    at SafeSubscriber.next (***/node_modules/rxjs/internal/Subscriber.js:143:22)
    at Subscriber._next (***/node_modules/rxjs/internal/Subscriber.js:89:26)
    at Subscriber.next (***/node_modules/rxjs/internal/Subscriber.js:66:18)
    at Subject.next (***/node_modules/rxjs/internal/Subject.js:60:25)
Emitted 'error' event on Socket instance at:
    at errorOrDestroy (internal/streams/destroy.js:128:12)
    at Socket.Writable.write (_stream_writable.js:324:5)
    at Logger.log (***/node_modules/concurrently/src/logger.js:107:27)
    [... lines matching original stack trace ...]
    at Subscriber.next (***/node_modules/rxjs/internal/Subscriber.js:66:18) {
  code: 'ERR_STREAM_DESTROYED'
}```

[gustavohenke]

Hey @idrise! Thanks for the report. Sounds like an easy fix.

[gustavohenke]

I was looking at this issue - while it does indeed seem like an easy fix, I'm not sure which feedback should concurrently provide. It'd be nice to print the usual "blabla exited with ", however everything goes to stdout. Switching to stderr for the final messages is probably not desirable. Maybe print some kind of error message instead?

[idrise]

I think a clean exit, killing the running processes makes the most sense IMHO. Could be that any throw ends up leaving processes hanging.