What is the problem?

Ray version and other system information (Python version, TensorFlow version, OS):
Ubuntu 18.04.4 LTS
Python 3.7.7
ray-0.9.0.dev0-cp37-cp37m-manylinux1_x86_64.whl

• For non integral values of x, eg. x = 3.2 in the code below, ray doesn't give a proper error message. Instead the worker dies. This also happens when x is not equal to the true num return values, eg. x = 5.
  ray.exceptions.RayWorkerError: The worker died unexpectedly while executing this task.
• For negative values of x, eg. x = -2 the kernel dies altogether.
  Aborted (core dumped)

Although one would not expect the user to enter such values. But still.

Reproduction

import ray

ray.init()

def f(a):
    return a, 2

x = <some_value>
g = ray.remote(num_return_vals = x)(f)

print(ray.get(g.remote(3)))

• I have verified my script runs in a clean environment and reproduces the issue.
• I have verified the issue also occurs with the latest wheels.