[Bug]: axes3d.py's _on_move() converts the roll angle to radians, but then passes it to view_init() as if it were still in degrees #28256

Bug summary

In \lib\mpl_toolkits\mplot3d\axes3d.py, _on_move() deals with rotation of 3d axes using the mouse. In line 1522, the roll angle (in degrees) is converted to radians:
roll = np.deg2rad(self.roll)
This roll in radians is used to calculate a new elevation and azimuth, like so:

    delev = -(dy/h)*180*np.cos(roll) + (dx/w)*180*np.sin(roll)
    dazim = -(dy/h)*180*np.sin(roll) - (dx/w)*180*np.cos(roll)
    elev = self.elev + delev
    azim = self.azim + dazim

A moment later, the view is updated:

    self.view_init(
        elev=elev,
        azim=azim,
        roll=roll,
        vertical_axis=vertical_axis,
        share=True,
    )

However, view_init() expects its parameters to be in degrees, not radians. As a consequence, the roll now diminishes by a factor pi/180 with every mouse movement. Not intended.

Code for reproduction

# Run the surface3d.py example, adding
ax.roll = 45
# It shows the plot, in the intended funny orientation (roll=45)
# Then move the mouse - you will see the orientation jump suddenly (to roll=0)

Actual outcome

The figure orientation has jumped to roll=0, after trying to rotate it only slightly by dragging the mouse:

Expected outcome

The figure is close to its original orientation (before dragging the mouse), at roll=45:

Additional information

Fix: add a line:
roll = self.roll
right after updating elev and azim (i.e., after line 1526).

Operating system

All, presumably; but I noticed it on Windows

Matplotlib Version

3.10.0.dev191+ge5af947d1b.d20240517

Matplotlib Backend

tkagg

Python version

Python 3.12.3

Jupyter version

No response

Installation

pip