Bug report

Bug summary

All's in the title.

One use case which showcases this is to prevent an artist from participating in autoscaling (in effect, a slightly hackish way to implement something like #15595):

ax = ...
dl = ax.dataLim.frozen()
ax.plot(...)
# restore old datalim, so that plot() does not participate in autoscale
ax.dataLim = dl

which works... except for log-scale, as that needs also copying minposx/minposy.

Code for reproduction

from pylab import *
plot([1, 2]); print(gca().dataLim.minposx, gca().dataLim.frozen().minposx)

Actual outcome

1.0 inf

Expected outcome

1.0 1.0

Matplotlib version

• Operating system: linux
• Matplotlib version (import matplotlib; print(matplotlib.__version__)): master (3.3.x)
• Matplotlib backend (print(matplotlib.get_backend())): any
• Python version: 3.9
• Jupyter version (if applicable):
• Other libraries: