Repository navigation
[Bug]: Plot window not shown in Mac OS with backend set to default MacOSX #26869
Description
Bug summary
Under version 3.8.0 the plot window does not appear when using MacOSX as the backend (default under Mac OS). This is not an issues with the previous 3.7.3 version or when backend is set to TkAgg.
Code for reproduction
import matplotlib
import matplotlib.pyplot as plt
import numpy as np
from matplotlib import cm
# matplotlib.use(backend="TkAgg")
plt.style.use("_mpl-gallery")
# Make data
X = np.arange(-5, 5, 0.25)
Y = np.arange(-5, 5, 0.25)
X, Y = np.meshgrid(X, Y)
R = np.sqrt(X**2 + Y**2)
Z = np.sin(R)
# Plot the surface
fig, ax = plt.subplots(subplot_kw={"projection": "3d"})
ax.plot_surface(X, Y, Z, vmin=Z.min() * 2, cmap=cm.Blues)
ax.set(xticklabels=[], yticklabels=[], zticklabels=[])
plt.show(block=False)
input("Hit [return] to end")
plt.close("all")Actual outcome
No plot is generated.
Expected outcome
Additional information
If the code line
matplotlib.use(backend="TkAgg")
is uncommented the code will work as expected.
Operating system
Mac OS 13.5.2
Matplotlib Version
3.8.0
Matplotlib Backend
MacOSX
Python version
3.11.5
Jupyter version
No response
Installation
conda
Activity
greglucas commentedon Sep 22, 2023
I can confirm this doesn't work anymore with
show(block=False), if you do block it works as expected. This came in via #25966 where we fixed sigint handling, but apparently introduced this.Specifically, it looks like it was the second cleanup commit removing unnecessary code: ec6717a
@anntzer I am guessing this may be related to the multiple signal tests being flaky as well... and this comment: #25966 (comment)
Nothing obvious jumps out to me at first glance other than the old versions specifically calling
CFRunLoopRun()within a while true block.ec6717a#diff-f8af2176752b51c29acf55ec5cc5d912f5459e19ed40559edcd78aaf7a6f6863L160
tacaswell commentedon Sep 22, 2023
Is
inputrunning an event loop for us?anntzer commentedon Sep 22, 2023
I don't really have the time to investigate this in depth right now, but I guess we could always revert the "remove unnecessary code" cleanup for now as a stopgap (if it turns out that it wasn't actually so unneeded?)
greglucas commentedon Oct 2, 2023
I think the root cause of this is that the cleanup removed the
PyOS_InputHookfunction that was in place and ran an event loop while waiting for the standard input. I just put up PR #26970 to try and re-implement some of that logic and things work as I'd expect, but happy to know if something still doesn't work or if there is a better way to do anything over there as well.tacaswell commentedon Oct 4, 2023
🤦🏻 Sorry I missed that when reviewing #26970 .