Without having looked at this too much in detail, it seems to come from blitting being used. Would blit=False show the expected behaviour?

@ImportanceOfBeingErnest Nice catch, blit=False shows the expected behaviour.

After more thought on it, I found a way to stop the animation which does not rely on the garbage collection of the Animation object (see code below for the record).

The behaviour described in the first message of this issue is not anymore an issue for me. Up to you to leave the issue open.

import numpy as np
import matplotlib
import matplotlib.pyplot as plt
from matplotlib.animation import FuncAnimation

fig, ax = plt.subplots()
xdata, ydata = [], []
(ln,) = plt.plot([], [], "ro")


def init():
    ax.set_xlim(0, 2 * np.pi)
    ax.set_ylim(-1, 1)
    return (ln,)


def update(frame):
    xdata.append(frame)
    ydata.append(np.sin(frame))
    ln.set_data(xdata, ydata)
    return (ln,)


def generate_data():
    stop_flag = False
    x = 0
    while not stop_flag:
        x += 2 * np.pi / 128
        stop_flag = yield x
    print("stopped")

data_generator = generate_data()

def stop():
    data_generator.send("STOP")

def onClose(event):
    stop()
    
def onPress(event):
    if event.key == " ":
        stop()
    
fig.canvas.mpl_connect("close_event", onClose)
fig.canvas.mpl_connect("key_press_event", onPress)
ani = FuncAnimation(fig, update, frames=data_generator, init_func=init, blit=False)
plt.show()

The culprit is indeed this line in _setup_blit:

This issue has been marked "inactive" because it has been 365 days since the last comment. If this issue is still present in recent Matplotlib releases, or the feature request is still wanted, please leave a comment and this label will be removed. If there are no updates in another 30 days, this issue will be automatically closed, but you are free to re-open or create a new issue if needed. We value issue reports, and this procedure is meant to help us resurface and prioritize issues that have not been addressed yet, not make them disappear. Thanks for your help!

mpl_connect should only be holding a weakref. This warrants more investigation.

This seems to be fixed as of master for me; @tacaswell can you confirm or infirm this?