Integrated repository, worktree, git stacker, and project manager
tests: widen ls concurrency threshold to 2.5x delay
`test_ls_runs_sources_concurrently` flaked under contended xdist
workers — elapsed times around 0.4s tripped the previous 2x = 0.4s
ceiling. Move to 2.5x; still well under the 3x serial fallback so a
real regression to serial would still trip the assertion.
Jordan Isaacs
(Apr 28, 2026, 6:41 PM UTC)
7cf96cf6
56cfb26c
+6
-3
+6
-3
tests/agent/test_ls.py
+6
-3
tests/agent/test_ls.py
···
203
203
started = time.monotonic()
204
204
asyncio.run(ls_mod.ls(pm_env, ["demo"], 5, parse_agents(None)))
205
205
elapsed = time.monotonic() - started
206
-
# Serial would be ~3*delay = 0.6s. Parallel should be ~delay = 0.2s.
207
-
# Pick a threshold comfortably below 2*delay to avoid flakes.
208
-
assert elapsed < 2 * delay
206
+
# Serial would be ~3*delay = 0.6s; truly parallel is ~delay = 0.2s.
207
+
# On contended xdist workers we've seen ~0.4-0.5s for the parallel run
208
+
# because thread scheduling slips a few hundred ms — still well under
209
+
# serial. Threshold sits halfway between (3*delay = 0.6s) and parallel
210
+
# so we catch a regression to serial without flaking on slow runners.
211
+
assert elapsed < 2.5 * delay
209
212
210
213
211
214
# --- pm agent ls --resume CLI ---