Skip to content

test(ask_user): de-flake two CI-timing failures (unblocks #237) #238

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AVADSA25 merged 1 commit into from
Jul 13, 2026
Merged

test(ask_user): de-flake two CI-timing failures (unblocks #237) #238

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 18 additions & 4 deletions tests/test_ask_user.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,7 +97,13 @@ def caller():
time.sleep(0.05)
assert qid is not None, "ask() never wrote a pending record"

# Verify notifications.json has a type="question" entry.
# Verify notifications.json has a type="question" entry. ask() writes the
# pending record and the notification as two separate writes, so under load
# the notification can lag the record we polled for above — poll for it too
# rather than assume it landed simultaneously (CI flake, 2026-07).
_nd = time.monotonic() + 2.0
while time.monotonic() < _nd and not notif_path.exists():
time.sleep(0.05)
assert notif_path.exists(), "notification not written"
notifs = json.loads(notif_path.read_text())
assert any(n.get("type") == "question" and n.get("pending_question_id") == qid
Expand Down Expand Up @@ -173,10 +179,18 @@ def test_deadline_timeout_returns_sentinel(temp_audit_log, temp_askuser_paths):
# Should be roughly 2s (polling loop checks every 1s); allow up to 4s.
assert 1.5 <= elapsed <= 4.5, f"deadline elapsed={elapsed:.2f}s"

# Audit: emit + timeout.
# Audit: emit + timeout. Scope to THIS question's id: a background-threaded
# ask() in a neighbouring test can finalize a late timeout into this test's
# audit log (the log path is read live at write-time), so a global count is
# flaky under CI load. Count only events for our own qid.
my_qid = json.loads(pq_path.read_text())["pending_questions"][0]["id"]

def _for_me(events):
return [e for e in events if e.get("extra", {}).get("pending_question_id") == my_qid]

recs = _records(temp_audit_log)
emits = _events_of(recs, codec_ask_user.ASKUSER_EVENT_EMIT)
timeouts = _events_of(recs, codec_ask_user.ASKUSER_EVENT_TIMEOUT)
emits = _for_me(_events_of(recs, codec_ask_user.ASKUSER_EVENT_EMIT))
timeouts = _for_me(_events_of(recs, codec_ask_user.ASKUSER_EVENT_TIMEOUT))
assert len(emits) == 1
assert len(timeouts) == 1
assert timeouts[0]["extra"]["reason"] == "deadline"
Expand Down