Python SDK test TestOperationCommands::test_operation_stderr_output fails

[dmi-feo]

tests/test_operations.py::TestOperationCommands::test_operation_stderr_output

        op = yt.run_map("echo -e -n 'someerrout' >&2; cat", input_table, output_table, sync=False, job_count=1)
        with set_config_option("operation_tracker/always_show_job_stderr", True):
            caplog.clear()                                                                                                                                                                                                                                op.wait()
        assert op.get_state() == "completed"
>       assert any("someerrout" in rec.message for rec in caplog.records)                                                                                                                                                                     E       assert False
E        +  where False = any(<generator object TestOperationCommands.test_operation_stderr_output.<locals>.<genexpr> at 0x7fbfd88c7d60>)
tests/test_operations.py:1033: AssertionError

It fails because caplog captures no records at all.

I've found two ways to make the test pass:

1. Enable propagate for Yt logger here. Having propagation enabled the root logger must start handling records as well as Yt logger. However, logging directly to the root logger here without propagation doesn't help, which is weird since the only logger records will be propagated to is the root one.
2. Configure logger the same way it is done for arcadia environment. It replaces Yt logger with the root one, but still it is unclear why the test fails when the root logger is used explicitly.

[dmi-feo]

By the way, is it okay that caplog.clear() is called after run_map? It seems it can wipe the information we wait for in case of too fast map operation.