Bug summary

A little tricky to explain, but see the outcomes below. Found while making tests for #28225

Code for reproduction

import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d.art3d import Poly3DCollection

theta = np.linspace(0, 2*np.pi, 50)

x1 = x2 = x3 = np.cos(theta)
y1 = y2 = y3 = np.sin(theta)
z1 = np.ones_like(theta) * 0
z2 = np.ones_like(theta) * 1
z3 = np.ones_like(theta) * 2

# flip direction of this face
x3 = x3[::-1]
y3 = y3[::-1]

verts = list(zip(x1, y1, z1)) + list(zip(x2, y2, z2)) + list(zip(x3, y3, z3))
verts = np.array([verts])
poly = Poly3DCollection(verts, alpha=.7, edgecolor='k')

fig = plt.figure()
ax = fig.add_subplot(projection='3d')
ax.add_collection3d(poly)
ax.auto_scale_xyz(verts[:, :, 0], verts[:, :, 1], verts[:, :, 2])
ax.set_aspect('equalxy')
plt.show()

Actual outcome

From this view, everything looks as expected

From this view, the overlapping area of the top and middle faces cancel out and show full transparency (bug).

From this view, the above is still happening, but the bottom face still shows properly even though it's in the intersection area.

The semi-transparent areas should really stack to a more opaque coloring as well, but that's a separate issue.

Expected outcome

If you don't reverse x3 and y3, the result looks as expected:

Additional information

No response

Operating system

No response

Matplotlib Version

latest main

Matplotlib Backend

No response

Python version

No response

Jupyter version

No response

Installation

git checkout