Normalizing Pixels to [0,1]

An attempt to scale uint8 pixels into the 0–1 range collapses everything to zeros.

import numpy as np

# img is uint8 with values 0..255
img_norm = img // 255
print(img_norm.min(), img_norm.max())  # expected ~0.0 and ~1.0

What is the bug in this normalization?