Files
rea/experiments/analysis-efficiency/joplin-cli-cancellation.py
180ba03bfd fix(electron): preserve cancellation across waits, state reads and cleanup
Fixes #4388. Refs #4027.

Observe cancellation during first-window waits and state reads; preserve cancellation and caller reasons before cleanup and retain partial observations when cleanup fails. Keep late cancellation and cleanup failure precedence.

Validation: 49 focused cases, static/docs checks, six controlled verifier-oracle cases and 39 current-head CI checks passed. Codex and Greptile found no remaining blocking issues. Real Joplin cancellation phases verified; complete Linux cleanup/first-signal CLI exit and packaged-runtime startup remain explicitly unverified/failed. Both verifiers now fail incomplete cleanup, including after main-process exit.

Co-authored-by: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
Co-authored-by: Adolanium <94890352+Adolanium@users.noreply.github.com>
2026-10-16 21:12:26 +08:00

124 lines
5.7 KiB
Python

#!/usr/bin/env python3
"""Verify real Joplin cancellation through REA's compiled public CLI.
The selected runtime and app run normally. Fresh profiles and raw CLI output
remain local; the printed receipt distinguishes cancellation from cleanup.
"""
import argparse
import json
from pathlib import Path
import signal
import subprocess
import tempfile
import time
def run(arguments):
profile = Path(tempfile.mkdtemp(prefix="joplin-cli-", dir=arguments.profile_root))
stdout_path = profile / "verifier-stdout.json"
stderr_path = profile / "verifier-stderr.txt"
request = {
"executable_path": str(arguments.executable.resolve()),
"application_path": str(arguments.application.resolve()),
"args": ["--no-sandbox", "--disable-gpu", f"--user-data-dir={profile}/chromium",
"--profile", str(profile)],
"actions": [{"step_id": "settle", "kind": "wait", "duration_ms": 60_000}],
}
started = time.monotonic()
cancelled = None
response_at = None
failure = None
signals = []
with stdout_path.open("wb") as output, stderr_path.open("wb") as errors:
process = subprocess.Popen(
[arguments.node, str(arguments.runtime_root.resolve() / "scripts/rea.mjs"),
"capture-electron-scenario", json.dumps(request), "--json"],
stdin=subprocess.DEVNULL, stdout=output, stderr=errors,
)
try:
readiness_deadline = started + 70
while not (profile / "log-main-process.txt").exists():
if process.poll() is not None:
raise RuntimeError(f"CLI exited before application startup: {process.returncode}")
if time.monotonic() > readiness_deadline:
raise TimeoutError("No real Joplin startup log within 70 seconds")
time.sleep(0.1)
# This only schedules cancellation. The returned action, not this
# delay, proves whether cancellation actually reached the wait.
try:
process.wait(timeout=3)
raise RuntimeError("CLI exited before the cancellation signal")
except subprocess.TimeoutExpired:
pass
cancelled = time.monotonic()
signals.append("request SIGTERM")
process.send_signal(signal.SIGTERM)
deadline = cancelled + 20
previous_size = 0
while process.poll() is None:
size = stdout_path.stat().st_size
if response_at is None and size != previous_size:
previous_size = size
try:
published = json.loads(stdout_path.read_text())
if isinstance(published, dict) and "code" in published:
response_at = time.monotonic()
except json.JSONDecodeError:
pass
if time.monotonic() > deadline:
raise TimeoutError("CLI did not exit within 20 seconds after request SIGTERM")
time.sleep(0.05)
except Exception as cause:
failure = cause
finally:
if process.poll() is None:
signals.append("verifier recovery SIGTERM")
process.terminate()
try:
process.wait(timeout=5)
except subprocess.TimeoutExpired:
signals.append("verifier recovery SIGKILL")
process.kill()
process.wait()
error = json.loads(stdout_path.read_text()) if stdout_path.stat().st_size else {}
partial = error.get("details", {}).get("partial_observation", {})
capture = partial.get("capture", {})
cancelled_action = any(action["status"] == "cancelled" for action in capture.get("actions", []))
receipt = {
"request": request,
"exit_code": process.returncode,
"elapsed_seconds": time.monotonic() - started,
"cancellation_to_observation_end_seconds": None if cancelled is None else time.monotonic() - cancelled,
"cancellation_to_response_seconds": None if response_at is None else response_at - cancelled,
"signals": signals,
"verifier_failure": None if failure is None else str(failure),
"code": error.get("code"), "category": error.get("category"),
"cancelled_action": cancelled_action,
"application": capture.get("application"),
"actions": capture.get("actions"), "coverage": capture.get("coverage"),
"stdout_bytes": stdout_path.stat().st_size,
"stderr_bytes": stderr_path.stat().st_size,
"stdout_path": str(stdout_path), "stderr_path": str(stderr_path),
"cleanup_verified": error.get("code") == "cancelled" and
capture.get("application", {}).get("cleanup") == "terminated-owned-process",
}
print(json.dumps(receipt))
if failure is not None:
raise failure
assert process.returncode == 143, "CLI must retain its SIGTERM exit convention"
assert partial.get("kind") == "electron-active-observation"
assert cancelled_action, "Startup alone does not establish action cancellation"
assert error.get("category") == "cancelled"
assert receipt["cleanup_verified"], "Cancellation alone does not establish complete provider cleanup"
if __name__ == "__main__":
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--node", required=True)
parser.add_argument("--runtime-root", type=Path, required=True)
parser.add_argument("--executable", type=Path, required=True)
parser.add_argument("--application", type=Path, required=True)
parser.add_argument("--profile-root", type=Path, required=True)
run(parser.parse_args())