Slice C: discovery deadlock, shared join deadline, shutdown hardening — v0.10.1

[jamesabel]

Summary

Slice C of the crash review: the hangs that read as crashes, and the shutdown-time native aborts. Two of the fixed defects were reproduced empirically during the review.

Fix	Before	After
GetTests joined before its queue was drained	Run preparation deadlocked forever past ~400–500 test modules (child can't exit until its queue feeder flushes; pipe blocks at ~64 KB) — all controls disabled, only escape was closing the window. Reproduced: 500+ items never exit	GetTests.collect(abort_event) drains while the child runs, honors abort (terminate + None)
PytestRunner.join(t) per-thread timeout	Wedged pool of N workers froze closeEvent (GUI thread) for (N+1)×t — minutes of "Not Responding" ending in a user kill	One shared deadline across the spin-up wait, every worker, and the runner thread
About tab QThread destroyed while running	Closing before the About data arrived (git can take seconds) → Qt fatal abort 0xC0000409, no Python traceback. Reproduced	About.shutdown() (quit + bounded 10 s wait), called from closeEvent
Prep thread outliving the window	Ignored the abort flag through PUT detection, a 120 s uninterruptible wind-down join, DB reads and RESUME copying; could start a runner after the window was gone and emit on a destroyed QObject (C++ use-after-free race)	Abort checked at every blocking phase (wind-down is a bounded abort-aware loop); emit guarded by shiboken6.isValid, with the prepared runner stopped instead of orphaned when nobody is left to adopt it
closeEvent re-entrancy	Its processEvents() calls could deliver a second close (nested teardown: second dialog, second stop) and full _update_tick passes — including reconcile_process_count against a stopping runner	_closing latch + refresh timer stopped before teardown

Tests

• test_collect_drains_large_result_sets_without_deadlock — 5000-item discovery completes in seconds (hung forever before)
• test_collect_aborts_wedged_discovery — pre-set abort terminates a wedged child promptly
• test_join_timeout_is_a_shared_deadline — join(1.0) with 3 busy workers returns in ~1 s, not 3+
• test_about_shutdown_stops_data_thread — thread finished, idempotent

Full suite: 501 passed. ruff clean; ty unchanged (13 pre-existing).

Behaviour changes

• join(t) is now a true overall bound — callers that relied on the accidental per-thread grace get less total waiting (that laxness was the freeze).
• Closing the window during run preparation aborts it promptly at the next phase boundary instead of letting it finish (and possibly start a runner) in the background.

🤖 Generated with Claude Code

https://claude.ai/code/session_01KEZ9TpT9VwBujBq1S3v2Jd

[codecov-commenter]

Codecov Report

❌ Patch coverage is 57.14286% with 24 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.78%. Comparing base (9511ae8) to head (6d1e916).

Files with missing lines	Patch %	Lines
src/pytest_fly/gui/run_tab/control_window.py	4.54%	21 Missing ⚠️
src/pytest_fly/gui/gui_main.py	71.42%	2 Missing ⚠️
src/pytest_fly/gui/about_tab/about.py	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #169      +/-   ##
==========================================
- Coverage   85.90%   85.78%   -0.13%     
==========================================
  Files          74       74              
  Lines        6224     6261      +37     
==========================================
+ Hits         5347     5371      +24     
- Misses        877      890      +13     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.