tensorflow: tf.maximum does not return nan when inputs contain nan

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): yes
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux CentOS 7
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): 1.3.0
• Python version: 2.7.13
• Bazel version (if compiling from source):
• CUDA/cuDNN version: 8.0/6.0
• GPU model and memory: Tesla K40m, 11439MiB
• Exact command to reproduce: python main.py

Describe the problem

tf.maximum(a, b) should return nan when a or b contain nan. However, it does not at some cases.

Source code / logs

main.py

import tensorflow as tf
import numpy as np

a = tf.placeholder(dtype=tf.float32)
max_a = tf.maximum(a, 1.)
with tf.Session():
    print max_a.eval(feed_dict={a: np.nan})

The output is:

1.0