`Axis3` animation: Non-continuous change of label positions when changing azimuth.

[manuelbb-upb]

Description

When changing the azimuth of an Axis3 to a multiple of π/2 in an animation, the axis labels do not move continuously, at the end they flip positions or disappear. See below animation: This is the script to make the animation:

using GLMakie

fig = Figure(;resolution=(500,500))
ax = Axis3(fig[1,1])

a0 = ax.azimuth[]
a1 = 3/2 * π

record(
    fig, joinpath(@__DIR__, "nonsmooth_axis3.gif"), 
    vcat(LinRange(a0,a1,100), fill(a1,50));
    fps = 25 
) do a
    ax.azimuth[] = a
end

Changing the elevation similarly seems to work smoothly.

System and Environment

Output of st in Julia 1.8.2:

  [e9467ef8] GLMakie v0.8.0
  [ee78f7c6] Makie v0.19.0

Output of uname -orv:

5.15.0-56-generic #62-Ubuntu SMP Tue Nov 22 19:54:14 UTC 2022 GNU/Linux

Output of sudo lshw -showrt -C cpu,memory,display:

H/W path           Device          Class          Description
=============================================================
/0/1                               memory         32GiB System Memory
/0/1/0                             memory         16GiB SODIMM DDR4 Synchronous 3200 MHz (0,3 ns)
/0/1/1                             memory         16GiB SODIMM DDR4 Synchronous 3200 MHz (0,3 ns)
/0/c                               memory         192KiB L1 cache
/0/d                               memory         128KiB L1 cache
/0/e                               memory         5MiB L2 cache
/0/f                               memory         12MiB L3 cache
/0/10                              processor      11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz
/0/11                              memory         128KiB BIOS
/0/100/2           /dev/fb0        display        TigerLake-LP GT2 [Iris Xe Graphics]
/0/100/14.2                        memory         RAM memory
/0/100/1d/0                        display        TU117M [GeForce MX450]

[jkrumbiegel]

They flip to the other side once the rotation is too far, seems like the side flipping doesn't perfectly coincide with the orientation flipping though. pi/2 is one of the breakpoints.

[ffreyer]

See also #1056