I can't seem to get the unit test that I want to work. I want to add the case to test_date_formatter_usetex

by adding the following:

@pytest.mark.parametrize('delta, expected', [
    ...
    (datetime.timedelta(minutes=10),
     [r'$\mathdefault{01\;00{:}%02d}$' % minu for minu in range(0, 11)]),
])
def test_date_formatter_usetex(delta, expected):
    ...

But I get an assertion error that the output is in a different format:

E       AssertionError: assert ['$\\mathdefa...00000}$', ...] == ['$\\mathdefa...{:}05}$', ...]
E         At index 0 diff: '$\\mathdefault{00{:}00{:}00.000000}$' != '$\\mathdefault{01\\;00{:}00}$'
E         Use -v to get the full diff

But when I run this script (just copied the function), it gives no assertion error.

import datetime

import matplotlib.dates as mdates

delta = datetime.timedelta(minutes=10)
expected = [r'$\mathdefault{01\;00{:}%02d}$' % minu for minu in range(0, 11)]

d1 = datetime.datetime(1990, 1, 1)
d2 = d1 + delta

locator = mdates.AutoDateLocator(interval_multiples=False)
locator.create_dummy_axis()
locator.axis.set_view_interval(mdates.date2num(d1), mdates.date2num(d2))

formatter = mdates.AutoDateFormatter(locator, usetex=True)
assert [formatter(loc) for loc in locator()] == expected

Am I doing anything wrong?

I can't look at details right now but tests are run under different defaults than current matplotlib. It's possible one of those has changed.

In classic.mplstyle, the minute formatter is date.autoformatter.minute : %H:%M:%S.%f whereas for modern matplotlib it is #date.autoformatter.minute: %d %H:%M

It would be great if you could update the test to modern matplotlib default formats, as we really should be testing those, and not the pre 2.0 ones ;-)