Bug summary

I have a very curious problem with num2date which I am not able to understand where it comes from. It seems that when running a code snippet (which is working as expected in a conda environment) within a container on a specific machine with time zone EET, num2date introduces a 2-hour offset from UTC associated to the time zone. The weird fact is that the time seems to be the same for both machines, so I do not understand where such extra two hours is applied. it is interesting also since the date2num works fine...

Unfortunately this is partially reproducible since this is happening within a singularity container that I cannot share easily. I am wondering if such potential issues with time zones are known to show up inside containers.

It might be a feature that I am not aware but I cannot understand how to control this.

Code for reproduction

from datetime import datetime, timezone
import matplotlib.dates as mdates
import time
print("Time:", time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()))
print("System timezone according to time module:", time.tzname)

original = datetime(1989, 1, 1, tzinfo=timezone.utc)
num = mdates.date2num(original)
converted = mdates.num2date(num, tz=timezone.utc)

print("Epoch:", mdates.get_epoch())
print("Original datetime:     ", original)
print("Converted to num:      ", num)
print("Back to datetime:      ", converted)

Actual outcome

On the container

Time: 2025-06-03 19:30:19
System timezone according to time module: ('EET', 'EEST')
Epoch: 1970-01-01T00:00:00
Original datetime:      1989-01-01 00:00:00+00:00
Converted to num:       6940.0
Back to datetime:       1988-12-31 22:00:00+00:00

Expected outcome

On the conda envrionment:

Time: 2025-06-03 19:30:23
System timezone according to time module: ('EET', 'EEST')
Epoch: 1970-01-01T00:00:00
Original datetime:      1989-01-01 00:00:00+00:00
Converted to num:       6940.0
Back to datetime:       1989-01-01 00:00:00+00:00

Additional information

Running matplotlib 3.10.3

Operating system

unix

Matplotlib Version

matplotlib 3.10.3

Matplotlib Backend

agg

Python version

3.12

Jupyter version

No response

Installation

conda