Bug#1126154: skimage: FTBFS in unstable: regression in test_deprecation_warnings_on_plugin_funcs

[Simon McVittie]

Source: skimage
Version: 0.25.2-3
Severity: important
Tags: ftbfs
Justification: fails to build from source (but built successfully in the past)

I tried building skimage as a smoke-test for a new version of 
python3-mesonpy, and saw a test failure which seems to be unrelated to 
the new python3-mesonpy. I'm reporting this as non-RC for now because I 
don't yet know whether it is reproducible in a clean environment - 
please bump the severity up to serious if you can reproduce the 
regression.

>=================================== FAILURES ===================================
>_________ test_deprecation_warnings_on_plugin_funcs[call_plugin-args1] _________
>
>func = <function call_plugin at 0x7f0d00ab4220>
>args = ['imread', '/build/reproducible-path/skimage-0.25.2/.pybuild/cpython3_3.13/build/skimage/data/camera.png']
>
>    @pytest.mark.parametrize(
>        ("func", "args"),
>        [
>            (io.use_plugin, ["imageio"]),
>            (io.call_plugin, ["imread", fetch("data/camera.png")]),
>            (io.plugin_info, ["imageio"]),
>            (io.plugin_order, tuple()),
>            (io.reset_plugins, tuple()),
>            (io.find_available_plugins, tuple()),
>            (getattr, [io, "available_plugins"]),
>        ],
>    )
>    def test_deprecation_warnings_on_plugin_funcs(func, args):
>        regex = ".*use `imageio` or other I/O packages directly.*"
>        with pytest.warns(FutureWarning, match=regex) as record:
>            func(*args)
>>       assert len(record) == 1
>E       assert 2 == 1
>E        +  where 2 = len(WarningsChecker(record=True))
>
>skimage/io/tests/test_plugin.py:92: AssertionError

It looks as though the test expects one deprecation warning, but sees 
two. Perhaps it could be relaxed to assert len(record) >= 1, which seems 
like it would be sufficient?

My local build was done in sbuild/unshare inside a trixie VM, with (I 
think) 4G RAM and 3 vCPUs. I've scheduled rebuilds on 
tests.reproducible-builds.org to see whether the same issue is 
reproducible there.

Thanks,
    smcv