Repository navigation
build(deps): keep PySide6 below 6.12 on Python < 3.12 - #125
Merged
Merged
Conversation
The `build (3.11, ubuntu-latest)` cell has hit its 20-minute timeout on
every run since the runner image moved to Ubuntu 26.04 (2026-10-29,
2026-11-01, 2026-11-02). The log shows what the timeout hid: xdist worker
gw0 died in tests/app/test_results_model.py with
Fatal Python error: none_dealloc: deallocating None: bug likely caused
by a refcount error in a C extension
and the session then hung on the replacement worker until the job was
killed, so the last line was a scan test that had in fact passed.
PySide6 6.12.0's Linux wheels drop a reference to ``None`` on ordinary
void Qt calls (upstream PYSIDE-3474). Tracking ``sys.getrefcount(None)``
per test on 3.11 shows the count sliding across the app suite until the
singleton is freed. On 3.12+ ``None`` is immortal (PEP 683), which is why
only the 3.10/3.11 cells were exposed and the 3.10 one passed by margin.
No 6.12.x release with a fix exists yet, so cap the app extra below 6.12
on Python < 3.12 (resolves to 6.11.3) and leave 3.12+ unchanged.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #125 +/- ##
==========================================
- Coverage 93.02% 92.99% -0.03%
==========================================
Files 43 43
Lines 4671 4671
==========================================
- Hits 4345 4344 -1
- Misses 326 327 +1
Flags with carried forward coverage won't be shown. Click here to find out more. 🚀 New features to boost your workflow:
|
JeanExtreme002
added a commit
that referenced
this pull request
Nov 3, 2026
Patch release carrying #125: the `app` extra now keeps PySide6 below 6.12 on Python < 3.12. The published 3.4.0 metadata still resolves to PySide6 6.12.0 there, whose Linux wheels leak references to ``None`` and abort the interpreter (upstream PYSIDE-3474), so the cap only protects users once it is on PyPI. Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
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
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.
Why
build (3.11, ubuntu-latest)has timed out at 20 minutes on every run since 2026-10-29 (runs 40818789463, 41273979268, 41355657134). The log looked like a hung memory scan, but the scan had passed. The real sequence:gw0aborts intests/app/test_results_model.py:gw2, and the session then sits on a futex until the job timeout. A diagnostic run on the same image confirmed this: all remaining tests had finished, every Python process was sleeping, and the crashed worker was a zombie.Root cause
PySide6 6.12.0's Linux wheels drop a reference to
Noneon ordinary void Qt calls (upstream PYSIDE-3474; reported by several projects, e.g. optiland/optiland#1069, eoyilmaz/displaycal-py3#998). Measuringsys.getrefcount(None)per test on 3.11 shows the count sliding across the app suite until the singleton is freed. On 3.12+Noneis immortal (PEP 683) so the leak is harmless, which is why only the 3.10/3.11 cells were exposed, and 3.10 passed by margin. The runner-image change only moved the baseline enough to tip 3.11 over.Fix
Cap the
appextra below 6.12 on Python < 3.12 (resolves to 6.11.3) and leave 3.12+ as is. No 6.12.x release with a fix exists yet; the comment inpyproject.tomlsays when to lift the cap.Verified locally on 3.11 with the pin: the app suite passes (234 tests) with PySide6 6.11.3.
Follow-up worth considering (not in this PR)
A crashed xdist worker currently turns into a 20-minute silent timeout.
--max-worker-restart=0would fail the job immediately with the crash in the log instead.🤖 Generated with Claude Code