lib/matplotlib/tests/test_backends_interactive.py::test_figure_leak_20490[time_mem0-MPLBACKEND=wxagg-BACKEND_DEPS=wx] has been failing for a while on CI now, just on the Tests / Python 3.10 on ubuntu-20.04 (pull_request) job. See e.g. this run for an example failure.
The trace is
______test_figure_leak_20490[time_mem0-MPLBACKEND=wxagg-BACKEND_DEPS=wx] ______
[gw1] linux--Python3.10.13/opt/hostedtoolcache/Python/3.10.13/x64/bin/pythonenv= {'BACKEND_DEPS': 'wx', 'MPLBACKEND': 'wxagg'}, time_mem= (0.0, 2000000)
request=<FixtureRequestfor<Functiontest_figure_leak_20490[time_mem0-MPLBACKEND=wxagg-BACKEND_DEPS=wx]>>@pytest.mark.skipif(sys.platform=="win32",reason="appveyor tests fail; gh-22988 suggests reworking")@pytest.mark.parametrize("env", _get_testable_interactive_backends())@pytest.mark.parametrize("time_mem", [(0.0, 2_000_000), (0.1, 30_000_000)])deftest_figure_leak_20490(env, time_mem, request):
pytest.importorskip("psutil", reason="psutil needed to run this test")
# We haven't yet directly identified the leaks so test with a memory growth# threshold.pause_time, acceptable_memory_leakage=time_memifenv["MPLBACKEND"] =="wx":
pytest.skip("wx backend is deprecated; tests failed on appveyor")
ifenv["MPLBACKEND"] =="macosx":
request.node.add_marker(pytest.mark.xfail(reason="macosx backend is leaky"))
ifenv["MPLBACKEND"] =="tkagg"andsys.platform=="darwin":
acceptable_memory_leakage+=11_000_000result=_run_helper(
_test_figure_leak, str(pause_time),
timeout=_test_timeout, extra_env=env)
growth=int(result.stdout)
>assertgrowth<=acceptable_memory_leakageEAssertionErrorlib/matplotlib/tests/test_backends_interactive.py:687: AssertionError
This test has been somewhat flaky/bandaged over for a while, but something recently has seemed to shift probabilities for wx... Not sure what that would be, though as wcPython has not seen a release recently as far as I can tell.
I will cross link to the other times this test has been touched that I can find easily here:
Activity
ksunden commented on Jan 11, 2024
This test has been somewhat flaky/bandaged over for a while, but something recently has seemed to shift probabilities for wx... Not sure what that would be, though as wcPython has not seen a release recently as far as I can tell.
I will cross link to the other times this test has been touched that I can find easily here:
#20490
#23384
#23094
#24711
#24397
#26680
We already fudge the numbers in some cases/ignore/xfail the test in others.
ksunden commented on Jan 11, 2024
I am also stochastically seeing this test fail locally (with wxagg) but with a slightly different presentation:
I'm not sure what is sending SIGTRAP, though I guess it could be memory related somewhere...