Weird pytest error: 'CompletionHtml' object has no attribute '_text_edit'

[juliangilbey]

While building qtconsole for Debian, we run the test suite in a clean environment. And with version 5.4.3, we're now often getting a very strange pytest error (the times at the beginning of the lines are a running clock):

 50s Testing with python3.11:
 50s ============================= test session starts ==============================
 50s platform linux -- Python 3.11.4, pytest-7.2.1, pluggy-1.0.0+repack
 50s PyQt5 5.15.9 -- Qt runtime 5.15.8 -- Qt compiled 5.15.8
 50s rootdir: /tmp/autopkgtest-lxc._jqfs5cq/downtmp/build.peT/src
 50s plugins: flaky-3.7.0, qt-4.2.0+repack
 50s collected 46 items
 50s 
109s qtconsole/tests/test_00_console_widget.py ................               [ 34%]
109s qtconsole/tests/test_ansi_code_processor.py ...........                  [ 58%]
110s qtconsole/tests/test_app.py .s                                           [ 63%]
111s qtconsole/tests/test_comms.py ..                                         [ 67%]
111s qtconsole/tests/test_completion_widget.py ....                           [ 76%]
111s qtconsole/tests/test_frontend_widget.py ..                               [ 80%]
111s qtconsole/tests/test_inprocess_kernel.py .                               [ 82%]
111s qtconsole/tests/test_jupyter_widget.py ...                               [ 89%]
111s qtconsole/tests/test_kill_ring.py ...                                    [ 95%]
111s qtconsole/tests/test_styles.py ..                                        [100%]
111s ===Flaky Test Report===
111s 
111s test_scroll[True] passed 1 out of the required 1 times. Success!
111s test_scroll[False] passed 1 out of the required 1 times. Success!
111s test_input passed 1 out of the required 1 times. Success!
111s test_debug passed 1 out of the required 1 times. Success!
111s test_input_and_print failed (14 runs remaining out of 15).
111s 	<class 'pytestqt.exceptions.TimeoutError'>
111s 	waitUntil timed out in 5000 milliseconds
111s 	[<TracebackEntry /tmp/autopkgtest-lxc._jqfs5cq/downtmp/build.peT/src/qtconsole/tests/test_00_console_widget.py:239>, <TracebackEntry /tmp/autopkgtest-lxc._jqfs5cq/downtmp/build.peT/src/qtconsole/tests/test_00_console_widget.py:223>]
111s test_input_and_print passed 1 out of the required 1 times. Success!
111s test_restart_after_kill passed 1 out of the required 1 times. Success!
111s 
111s ===End Flaky Test Report===
111s 
111s =================== 45 passed, 1 skipped in 60.96s (0:01:00) ===================
111s Traceback (most recent call last):
111s   File "/tmp/autopkgtest-lxc._jqfs5cq/downtmp/build.peT/src/qtconsole/completion_html.py", line 149, in eventFilter
111s     if obj == self._text_edit:
111s               ^^^^^^^^^^^^^^^
111s AttributeError: 'CompletionHtml' object has no attribute '_text_edit'
111s Aborted

It looks like a qtbot.waitUntil timeout (from the TracebackEntry), but then there is the very strange error at the end about the _text_edit attribute being missing.

Do you have any idea what might be causing this or how to fix it?

[juliangilbey]

I just realised that the qbot.waitUntil timeout is in a flaky report, and is fine. But it is the weird _text_edit issue which is frequently happening, even when the test_input_and_print call is successful first time.

[juliangilbey]

Update: I have located the source of the issue (though I don't understand why it is breaking); it is the test in qtconsole/tests/test_inprocess_kernel.py which causes the error. Running this test under unittest rather than pytest passes with no issues.

[rayosborn]

Testing is not really my area of expertise, but the tests were set up as unittests so I'm not completely sure why you want to use pytest. In my clone of the repository, test_inprocess_kernel.py gives no error.

[juliangilbey]

Hi @rayosborn, some of the test files contain only pytest tests, some contain only unittest tests and some contain a mixture. The GitHub workflow runs all the tests using pytest:

qtconsole/.github/workflows/linux-tests.yml

Lines 73 to 78 in d700870

	- name: Run tests
	shell: bash -l {0}
	run: xvfb-run --auto-servernum pytest -vv -s --full-trace --color=yes --cov=qtconsole qtconsole
	env:
	QT_API: ${{ matrix.QT_LIB }}
	PYTEST_QT_API: ${{ matrix.QT_LIB }}

But for some strange reason, this caused the error described above on a Debian system.