gh-150387: Fix hang in test_run_failed_script_live on slow buildbots

pablogsal · pablogsal · commit 0f17e0b9ba0e · 2026-05-25T15:40:15.000+01:00
The test relied on the failing script (``1/0``) crashing fast enough that
fewer than ``MIN_SAMPLES_FOR_TUI`` (200) samples were collected, which would
trigger the early-return path in ``sample.sample_live`` and bypass the TUI
input loop. On slow buildbots (ASan, TraceRefs, NoGIL) the interpreter
takes long enough to bootstrap and reach the ``1/0`` that the profiler
easily collects &gt;=200 samples, pushing execution into the
``while collector.running:`` loop. Because the mock only queued a final
``q`` keystroke when ``n_times &gt;= 500`` (and the failed-script test used
``n_times=200``), nothing ever set ``running`` to ``False`` and the test
hung until the regrtest timeout.

Always queue the final ``q`` so the live TUI loop exits regardless of how
many samples were collected.
diff --git a/Lib/test/test_profiling/test_sampling_profiler/test_live_collector_ui.py b/Lib/test/test_profiling/test_sampling_profiler/test_live_collector_ui.py
@@ -835,8 +835,7 @@ def mock_init_curses_side_effect(self, n_times, mock_self, stdscr):
         # still failing
         for _ in range(n_times):
             mock_self.display.simulate_input(-1)
-        if n_times >= 500:
-            mock_self.display.simulate_input(ord('q'))
+        mock_self.display.simulate_input(ord('q'))
 
     def test_run_failed_module_live(self):
         """Test that running a existing module that fails exits with clean error."""
diff --git a/Misc/NEWS.d/next/Tests/2026-05-25-15-39-53.gh-issue-150387.yzZ7jq.rst b/Misc/NEWS.d/next/Tests/2026-05-25-15-39-53.gh-issue-150387.yzZ7jq.rst
@@ -0,0 +1,5 @@
+Fix hang in
+``test.test_profiling.test_sampling_profiler.test_live_collector_ui.TestLiveModeErrors.test_run_failed_script_live``
+on slow buildbots. The test now always queues a final ``q`` keystroke so the
+live TUI loop exits even when the profiler collects enough samples to enter
+the post-finished input loop.
