fix: monitor worker non-blocking + test fixes

- core/monitor.py: add fast metrics script (0.3s vs 1.4s sleep), collect() now defaults to fast=True for responsive UI - ui/workers.py: MonitorWorker changed from loop mode to single-shot (QTimer re-arms next cycle on finished signal, avoids paramiko blocking in infinite loop) - ui/widgets.py: remove debug print from _kick_one_sample - test_monitor_nonblock.py: fix sample counting by wrapping _on_sample instead of disconnecting signals (old approach missed new workers) - test_process_monitor.py: assertions account for MAX_RENDER_ROWS=200, fix MEM% column index (4 not 3) in sort test
2026-07-28 22:57:15 +08:00
parent 5e69b1c3e2
commit 5d83348db3
5 changed files with 341 additions and 33 deletions
@@ -0,0 +1,142 @@
+"""
+监控 UI 不卡死测试：
+- 启动监控
+- 5 秒内应该完成 ≥2 次采集（说明没卡死）
+- worker 不会重复启动
+- 关闭/重新打开工作正常
+- _on_refresh_now 不打乱 timer
+"""
+import os
+import sys
+import time
+import threading
+
+os.environ["QT_QPA_PLATFORM"] = "offscreen"
+sys.path.insert(0, os.path.dirname(__file__))
+
+from PyQt5.QtCore import QTimer
+from PyQt5.QtWidgets import QApplication
+
+from core.ssh_client import SSHConnection
+from core.monitor import SystemMonitor
+from core.manager import ConnectionManager
+from ui.widgets import MonitorPanel
+
+
+def find_local_ssh():
+    for pwd in ("sshclient_test_pwd_2026", "testpass", ""):
+        c = SSHConnection("127.0.0.1", 22, "root", pwd, timeout=5)
+        ok, _ = c.connect()
+        if ok:
+            return c
+    return None
+
+
+def main():
+    app = QApplication(sys.argv)
+    print("[1/6] 连接本机 SSH")
+    conn = find_local_ssh()
+    if not conn:
+        print("  ⚠ 本机 SSH 不可用，跳过")
+        return
+    print(f"  ✓ {conn.username}@{conn.host}")
+
+    print("[2/6] 创建 MonitorPanel + 启动监控")
+    mgr = ConnectionManager()
+    mgr.connections["fake"] = conn
+    panel = MonitorPanel(mgr)
+    panel.set_host("fake")
+    panel.btn_toggle.setChecked(True)  # 等价于点「开始监控」
+    app.processEvents()
+    assert panel.btn_toggle.isChecked()
+    assert panel._monitor_conn is not None
+    print(f"  ✓ 监控启动, interval={panel._monitor_interval_ms}ms")
+
+    print("[3/6] 5 秒内累计采集次数（要求 ≥2 次且不卡死）")
+    # 改用更短的 interval 让测试在 5s 内能采到 2 次以上
+    panel._monitor_interval_ms = 1000
+    # 用 wrapper 替换 _on_sample 来计数。
+    # 关键：_kick_one_sample 里每次创建新 worker 时执行
+    #   self.worker.sample_ready.connect(self._on_sample)
+    # 此处 self._on_sample 在连接时求值，所以新 worker 会连到 wrapper。
+    # 旧的断开/重连方式只作用于当前 worker，新 worker 不经过 counting。
+    orig_on_sample = panel._on_sample
+    samples = [0]
+    def counting(m):
+        samples[0] += 1
+        orig_on_sample(m)
+    panel._on_sample = counting
+    # 重启 timer 到 1s（如果 worker #1 还在跑，等它结束后 _on_worker_finished
+    # 会创建/重启 timer，此时 _monitor_interval_ms 已是 1000）
+    if getattr(panel, "_monitor_timer", None):
+        panel._monitor_timer.stop()
+        panel._monitor_timer.start(1000)
+    # 立即触发一次采集（如果 worker #1 已结束的话）
+    panel._kick_one_sample()
+    # 跑 5 秒（interval 1s + ~0.4s 采集耗时，应该能采 ~3 次）
+    deadline = time.time() + 5.0
+    while time.time() < deadline:
+        app.processEvents()
+        time.sleep(0.05)
+    wr = panel.worker.isRunning() if panel.worker else 'N/A'
+    timer = getattr(panel, '_monitor_timer', None)
+    print(f"  调试: samples={samples[0]} worker.isRunning={wr} timer.isActive={timer.isActive() if timer else 'N/A'}")
+    assert samples[0] >= 2, f"5s 内只采了 {samples[0]} 次，太少（卡死了？）"
+    print(f"  ✓ 5s 内采集次数: {samples[0]} (期望 ≥2)")
+
+    print("[4/6] 验证 worker 不会重复启动")
+    assert panel.worker is not None
+    # worker 刚跑完应在 finished 状态
+    deadline = time.time() + 2
+    while panel.worker and panel.worker.isRunning() and time.time() < deadline:
+        app.processEvents()
+        time.sleep(0.05)
+    assert not panel.worker.isRunning(), "上次的 worker 应该在 finished 状态"
+    # 现在再点 refresh_now
+    samples[0] = 0
+    panel._on_refresh_now()
+    app.processEvents()
+    time.sleep(0.5)
+    app.processEvents()
+    assert samples[0] >= 1 or panel.worker.isRunning(), "refresh_now 后没采集到"
+    print(f"  ✓ 立即刷新后采集 1 次（worker isRunning={panel.worker.isRunning() if panel.worker else 'N/A'}）")
+    # 恢复原始 _on_sample
+    panel._on_sample = orig_on_sample
+
+    print("[5/6] 关闭监控（停止 worker）")
+    panel.btn_toggle.setChecked(False)
+    app.processEvents()
+    # 等 worker 退出
+    if panel.worker:
+        deadline = time.time() + 2
+        while panel.worker.isRunning() and time.time() < deadline:
+            app.processEvents()
+            time.sleep(0.05)
+        assert not panel.worker.isRunning(), "关闭后 worker 还在跑"
+    assert panel._monitor_timer is None, "关闭后 timer 还在"
+    print(f"  ✓ 监控关闭 (worker stopped, timer cleared)")
+
+    print("[6/6] 重新打开监控应能正常恢复")
+    panel.btn_toggle.setChecked(True)
+    app.processEvents()
+    assert panel._monitor_conn is not None
+    assert panel._monitor_interval_ms > 0
+    # 等 worker 跑完
+    deadline = time.time() + 3
+    while time.time() < deadline:
+        app.processEvents()
+        if panel.worker and not panel.worker.isRunning() and panel._proc_data:
+            break
+        time.sleep(0.05)
+    assert panel._proc_data, "重开后没采到数据"
+    print(f"  ✓ 重开正常，进程数={len(panel._proc_data)}")
+
+    # 清理
+    panel.btn_toggle.setChecked(False)
+    app.processEvents()
+    conn.disconnect()
+    print("\n监控 UI 不卡死测试通过 ✓")
+
+
+if __name__ == "__main__":
+    main()