strange result of rayTest in pybullet

[HiroIshida]

I am trying rayTest from pybullet, and having some strange results. It would be really nice if someone points out my mistakes, or takes this as a bug-report if it's a bug. The following python script performs a rayTest against a small_sphere.

import pybullet as p
import pybullet_data
import numpy as np

CLIENT = p.connect(p.GUI)
p.configureDebugVisualizer(p.COV_ENABLE_GUI, 0)
p.configureDebugVisualizer(p.COV_ENABLE_SHADOWS, 0)
p.setAdditionalSearchPath(pybullet_data.getDataPath()) #used by loadURDF

def test_smallsphere(pos_): 
    pos = np.array(pos_)
    ss = p.loadURDF("sphere_small.urdf")
    p.resetBasePositionAndOrientation(ss, pos, [0, 0, 0, 1])
    margin = np.array([0.3, 0.0, 0.0])
    p_start = pos - margin
    p_end = pos + margin
    p.addUserDebugLine(p_start, p_end, [1, 0, 0])
    result = p.rayTest(p_start, p_end)
    return result

result = test_smallsphere([0.0, 0, 0.000])
print(result)

This script works and shows a reasonable result:

((0, -1, 0.45000000000000007, (-0.02999999999999997, 0.0, 0.0), (-1.0, 0.0, 0.0)),)

Something strange is when I change the position of the sphere, for example when I change the argument to result = test_smallsphere([0.0, 0, 0.1]), no ray-collision is detected, In other words, I got result:

((-1, -1, 1.0, (0.0, 0.0, 0.0), (0.0, 0.0, 0.0)),)

This result is strange. I don't change relative geometrical relation between the ray and the small sphere, so it suppose to collide. Just for double check, the above script visualizes the line between p_start and p_end and I confirmed that it actually penetrate the sphere.

I use ubuntu 16.04 and the pybullet library is installed via pip(python 3.7.7). When I import pybullet inside python3, pybullet build time: Jun 19 2020 04:02:40 appears. This version seems newer and I guess the topics discussed in #1947 are probably unrelated.