Quarkus:dev is unkillable if StartupEvent has infinite loop

To reproduce, put this in a quarkus project:

@ApplicationScoped
public class FactorioDataGenerator {

    public void generateDemoData(@Observes StartupEvent startupEvent) {
           while (true) {
               // If needed Do something to avoid this code getting optimized away.
           }
    }
}

and then run mvn quarkus:dev in the terminal (I used the IntelliJ terminal). As your computer starts acting wierd, hit ctrl-c, which will give you back your terminal prompt. Then look at your OS task navigator. Notice that a process is still running at full CPU usage and it refused to die: that ctrl-c didn’t kill it.

[stuartwdouglas]

System.exit(1) won’t exit if there is a non-daemon thread stuck in an infinite loop. Your only options are Thread.stop() which is deprecated for very good reason, and kill -9.

We could use a shutdown hook in Maven to wait X seconds then kill the process I guess. It’s only dev mode, so that might be ok.

[stuartwdouglas]

Btw, just killing the process right away (without a grace period) on CTRL+C doesn’t seem like a good idea since that would prevent shutdown hooks from running (e.g. see #12577).

Exactly, from a production point of view you just can’t really do this. You can have an infinite loop in other threads as well, you just can’t guard against this.

I am going to close this, basically you can only have one of the following guaranteed:

a) That all shutdown code that should run is being run (i.e. that you can do a clean shutdown) b) That shutdown completes in a given time period

I think it is more important to guarantee a) than to try and do b) to guard against developer mistakes.