test: deterministically tear down example widgets - #743
Open
larsoner wants to merge 4 commits into
Open
Conversation
The namespace from runpy.run_path is discarded, so widgets created by
an example are garbage-collected at an arbitrary later point during
event processing. A still-visible widget whose C++ side is deleted
mid-event-loop can receive a paintEvent during deletion, which PyQt5
escalates to a process abort:
RuntimeError: wrapped C/C++ object of type QRangeSlider has been
deleted
as seen intermittently on the windows-latest (3.12) pyqt5 job in
test_example[demo_widgets/range_slider.py] (e.g. the scheduled main
run on 2026-08-22, and both pyapp-kit#741 and pyapp-kit#742).
Close (hide) all top-level widgets first so nothing can paint, then
deleteLater() + sendPostedEvents(None, DeferredDelete) to destroy the
native widgets (and stop their timers) while hidden -- note that
processEvents() alone does NOT dispatch DeferredDelete events -- then
gc.collect() for anything left over.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
pytest-qt 4.4.0 is incompatible with recent PySide6: qtbot.waitSignal fails with 'SystemError: <method disconnect of SignalInstance> returned a result with an exception set' and a TimeoutError in test_parent_changed (reproduced locally with PySide6 6.11.2 + pytest-qt 4.4.0; fixed by pytest-qt 4.5.0, see pytest-dev/pytest-qt#552). This is what has been failing every pyside6 CI job (and the scheduled runs on main) since at least mid-July. pytest-qt 4.5 dropped PySide2, so the pyside2 dependency group keeps the 4.4.0 pin via a separate test-qt-legacy group, declared as a uv conflict so universal resolution still succeeds. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
napari main now requires python >=3.11, so the job fails at install
time ('Package napari requires a different Python: 3.10.21 not in
>=3.11'). The pinned v0.4.19.post1 entry stays on 3.10.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
hatchling 1.32.0 (2026-08-11) bumped the default core metadata version
to 2.5, which the twine run by build-and-inspect-python-package rejects
('InvalidDistribution: 2.5 is not a valid metadata version'). Nobody
noticed because build-and-inspect needs the test job, which has been
red since mid-July (see previous commit). Verified locally: with this
setting both the wheel and sdist emit Metadata-Version 2.4 and pass
'twine check --strict'.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Author
|
... let me know if you want me to add a "open an issue when the scheduled one fails, if one doesn't already exist" here or in another PR. I find them useful in other repos, otherwise I miss scheduled run failures (but not sure if it would help here, depends on maintainer availability etc.) |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
CIs seem to be failing in all PRs. Based on (painful) experience with PyVistaQt, mne-qt-browser, etc. a reasonable fix is to make sure windows are closed and torn down at the end of every test (rather than "whenever gc.collect() decides to run"). Otherwise, a still-visible widget whose C++ side is deleted mid-event-loop can receive a paintEvent during deletion, which leads to a segfault.
This should hopefully fix CIs. Changes drafted by Claude Fable 5 but reviewed / iterated / understood be me.