loop-uh
fe37d0e7a3
Some checks failed
Windows project source guards / test (push) Failing after 39s
Замер на win10: у «шумного» ядра каждая строка лога с новым временем, id соединения или портом становилась отдельной записью журнала, а весь журнал копировался в GUI на каждую строку и переформатировался — CPU и память GUI росли с аптаймом (при 15 тыс. записей +0,19 ядра, +33 МБ). - RuntimeErrorJournal: записи группируются без учёта чисел (последний исходный текст сохраняется), не больше 300 записей — вытесняется та, что дольше всех не повторялась; порядок снимка прежний. - Снимок журнала уходит в GUI не чаще раза в секунду; вкладка «Ошибки ядер» форматирует записи только когда открыта. - Сбор трафика по процессам: вместо набора всех id соединений за сессию — владелец живых соединений (удаляется при закрытии) и счётчик на процесс.
164 lines
10 KiB
Python
164 lines
10 KiB
Python
import json
|
|
from pathlib import Path
|
|
import tempfile
|
|
from types import SimpleNamespace
|
|
import unittest
|
|
from unittest.mock import Mock, patch
|
|
import zipfile
|
|
|
|
from PyQt6.QtCore import QByteArray, QCoreApplication, QProcess
|
|
|
|
from xray_fluent.application.controller import AppController
|
|
from xray_fluent.engines.hysteria.runtime_contract import HysteriaTransitionContract, classify_hysteria_failure
|
|
from xray_fluent.diagnostics.runtime_errors import RuntimeErrorJournal, core_failure
|
|
from xray_fluent.diagnostics.export import export_diagnostics
|
|
from xray_fluent.engines.hysteria.manager import HysteriaManager
|
|
from xray_fluent.profiles.models import AppState
|
|
from xray_fluent.importer.link_parser import parse_single
|
|
|
|
|
|
_APP = QCoreApplication.instance() or QCoreApplication([])
|
|
|
|
|
|
class RuntimeErrorTests(unittest.TestCase):
|
|
def test_loopback_reset_from_process_pipe_does_not_abort_ready_session(self):
|
|
line = 'WARN SOCKS5 TCP error {"addr":"127.0.0.1:28327","reqAddr":"dns.quad9.net:443","error":"read tcp4 127.0.0.1:11809->127.0.0.1:28327: wsarecv: An existing connection was forcibly closed by the remote host."}'
|
|
for running in (False, True):
|
|
with self.subTest(running=running):
|
|
manager = HysteriaManager()
|
|
manager._running = running
|
|
manager._process_generation = 8
|
|
controller = Mock()
|
|
controller.hysteria = manager
|
|
controller._hysteria_active_generation = 8
|
|
journal = RuntimeErrorJournal()
|
|
manager.log_received.connect(lambda message: journal.record(core_failure('hysteria', 'runtime', message)))
|
|
manager.failure.connect(lambda code, message, generation:
|
|
AppController._on_hysteria_failure(controller, manager, code, message, generation))
|
|
with patch.object(manager._process, 'readAllStandardOutput', return_value=QByteArray((line + '\n').encode())):
|
|
manager._on_ready_read()
|
|
self.assertFalse(manager._failure_reported)
|
|
self.assertIsNone(manager.last_failure_code)
|
|
controller.singbox.stop.assert_not_called()
|
|
controller.singbox.request_stop.assert_not_called()
|
|
controller._request_transition.assert_not_called()
|
|
with patch.object(manager._process, 'state', return_value=QProcess.ProcessState.Running), \
|
|
patch.object(manager, '_probe_remote_endpoint', return_value=None):
|
|
self.assertTrue(manager._wait_until_remote_ready(11809, username='', password='', timeout=1))
|
|
record = journal.snapshot()[0].failure
|
|
self.assertEqual(record.code, 'LOCAL_CLIENT_CONNECTION_CLOSED')
|
|
self.assertEqual(record.action, 'record_only')
|
|
self.assertIn(line, record.message)
|
|
self.assertIn('stage=local_client', record.message)
|
|
|
|
def test_loopback_diagnostic_does_not_hide_remote_or_process_failure(self):
|
|
for address in ('127.0.0.1:11809->127.0.0.1:20000', '[::1]:11809->[::1]:20000'):
|
|
message = f'write tcp {address}: write: broken pipe'
|
|
self.assertEqual(core_failure('hysteria', 'runtime', message).action, 'record_only')
|
|
self.assertIsNone(classify_hysteria_failure(message))
|
|
self.assertEqual(classify_hysteria_failure(message, process_exited=True).value, 'LOCAL_PROCESS_EXITED')
|
|
for text, expected in (
|
|
('connect error: connection refused', 'TARGET_CONNECTION_REFUSED'),
|
|
('read tcp4 127.0.0.1:11809->198.51.100.3:443: connection reset by peer', 'TARGET_CONNECTION_CLOSED'),
|
|
('connect error: An existing connection was forcibly closed by the remote host.', 'TARGET_CONNECTION_CLOSED'),
|
|
('connect error: timeout: no recent network activity', 'TARGET_NETWORK_TIMEOUT'),
|
|
('connect error: INTERNAL_ERROR (local): no certificate matches the pinned hash', 'TARGET_PIN_MISMATCH'),
|
|
):
|
|
self.assertEqual(classify_hysteria_failure(text).value, expected)
|
|
|
|
def test_manager_security_escalation_reaches_controller_after_timeout(self):
|
|
current = parse_single('hy2://current@one.example:443/#current')
|
|
replacement = parse_single('hy2://replacement@two.example:443/#replacement')
|
|
manager = HysteriaManager()
|
|
manager._running = True
|
|
manager._process_generation = 7
|
|
controller = Mock()
|
|
controller.hysteria = manager
|
|
controller._hysteria_active_generation = 7
|
|
controller._hysteria_contract = HysteriaTransitionContract()
|
|
controller._hysteria_contract.begin(3, current.id, 'official_hysteria_sidecar')
|
|
controller.connected = True
|
|
controller._disconnecting = False
|
|
controller._desired_connected = True
|
|
controller._hysteria_recovery_active = False
|
|
controller._hysteria_cooldown_until = {}
|
|
controller._hysteria_failure_episode_id = 0
|
|
controller.selected_node = current
|
|
controller.state = AppState(nodes=[current, replacement], selected_node_id=current.id)
|
|
journal = RuntimeErrorJournal()
|
|
manager.log_received.connect(lambda message: journal.record(core_failure('hysteria', 'runtime', message)))
|
|
manager.failure.connect(lambda code, message, generation:
|
|
AppController._on_hysteria_failure(controller, manager, code, message, generation))
|
|
manager._emit_process_line('WARN connect error: timeout: no recent network activity')
|
|
manager._emit_process_line('WARN x509: certificate signed by unknown authority')
|
|
manager._emit_process_line('WARN connect error: timeout: no recent network activity')
|
|
self.assertEqual(controller._request_transition.call_count, 1)
|
|
self.assertFalse(controller._desired_connected)
|
|
self.assertEqual(controller._hysteria_last_failure_code.value, 'TARGET_TLS_UNKNOWN_AUTHORITY')
|
|
self.assertIn('unknown authority', controller._set_connection_status.call_args.args[1])
|
|
self.assertEqual([record.failure.code for record in journal.snapshot()],
|
|
['TARGET_NETWORK_TIMEOUT', 'TARGET_TLS_UNKNOWN_AUTHORITY'])
|
|
self.assertEqual(controller.state.selected_node_id, current.id)
|
|
self.assertEqual(controller._hysteria_contract.session.failure_episode_id, 1)
|
|
|
|
def test_unknown_error_is_kept_with_no_invented_diagnosis(self):
|
|
message = 'ERROR upstream returned odd status 781: untouched details'
|
|
failure = core_failure('xray', 'dial', message)
|
|
self.assertEqual(failure.message, message)
|
|
self.assertEqual(failure.code, 'CORE_UNCLASSIFIED')
|
|
self.assertEqual(failure.target_id, '')
|
|
|
|
def test_journal_export_is_independent_of_traffic_limit(self):
|
|
# Журнал не обрезается лимитом лога трафика: все РАЗНЫЕ ошибки (в
|
|
# пределах MAX_JOURNAL_RECORDS) доходят до экспорта, повторы считаются.
|
|
# Строки, различающиеся только числами (время, id, порты), — одна запись.
|
|
journal = RuntimeErrorJournal()
|
|
kinds = [f"ERROR evidence {chr(97 + i % 26)}{chr(97 + i // 26)}" for i in range(250)]
|
|
for kind in kinds:
|
|
journal.record(core_failure('sing-box', 'dial', kind))
|
|
journal.record(core_failure('sing-box', 'dial', kinds[0]))
|
|
with tempfile.TemporaryDirectory() as directory:
|
|
path = export_diagnostics(Path(directory) / 'report.zip', AppState(), [],
|
|
runtime_errors=journal.snapshot())
|
|
with zipfile.ZipFile(path) as archive:
|
|
records = json.loads(archive.read('runtime_errors.json'))
|
|
self.assertEqual(len(records), 250)
|
|
self.assertEqual(records[0]['occurrences'], 2)
|
|
self.assertEqual(records[0]['failure']['message'], kinds[0])
|
|
|
|
def test_manager_log_reaches_security_policy_after_ready(self):
|
|
for line, code in (
|
|
('ERROR x509: certificate signed by unknown authority', 'TARGET_TLS_UNKNOWN_AUTHORITY'),
|
|
('ERROR server certificate SHA-256 mismatch', 'TARGET_PIN_MISMATCH'),
|
|
('ERROR authentication error: invalid user', 'TARGET_AUTH_REJECTED'),
|
|
('WARN obfs rejected', 'TARGET_OBFS_REJECTED'),
|
|
):
|
|
with self.subTest(line=line), patch.object(HysteriaManager, '_cleanup_config'), \
|
|
patch.object(HysteriaManager, '_cleanup_stale_generation_configs'):
|
|
manager = HysteriaManager()
|
|
manager._running = True
|
|
manager._process_generation = 7
|
|
contract = HysteriaTransitionContract()
|
|
contract.begin(3, 'node', 'official_hysteria_sidecar')
|
|
controller = SimpleNamespace(
|
|
hysteria=manager, _hysteria_active_generation=7,
|
|
connected=True, _disconnecting=False, _desired_connected=True,
|
|
_hysteria_contract=contract, singbox=Mock(is_running=True),
|
|
_log=Mock(), _set_connection_status=Mock(), _request_transition=Mock(),
|
|
)
|
|
journal = RuntimeErrorJournal()
|
|
manager.log_received.connect(lambda message: journal.record(core_failure('hysteria', 'runtime', message)))
|
|
manager.failure.connect(lambda failure_code, message, generation:
|
|
AppController._on_hysteria_failure(controller, manager, failure_code, message, generation))
|
|
manager._emit_process_line(line)
|
|
controller.singbox.request_stop.assert_called_once_with(expected=True)
|
|
controller.singbox.stop.assert_not_called()
|
|
controller._request_transition.assert_not_called()
|
|
self.assertFalse(controller._desired_connected)
|
|
self.assertEqual(controller._hysteria_last_failure_code.value, code)
|
|
self.assertIn(line, controller._set_connection_status.call_args.args[1])
|
|
self.assertEqual(journal.snapshot()[0].failure.code, code)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|