mirror of
https://github.com/shy3130/tick-stock-panel.git
synced 2026-11-02 10:26:13 +00:00
shy3130
a1301e239a
- 使用系统文件锁替代 PID 存活判断，崩溃后由系统释放锁并避免删除重建竞态 - 将早期预检异常统一交给启动错误提示，保证退出时释放已取得的锁 - GUI 就绪后显式显示主窗口，避免继承隐藏启动状态 - 移除安装包中重复的安装后启动条目 补充跨进程单实例锁、模拟 Windows 锁异常、早期启动失败、窗口显示与安装配置测试。 Windows 真机安装与窗口行为未在本机验证。 Signed-off-by: shy3130 <shy3130@qq.com>
471 lines
18 KiB
Python
471 lines
18 KiB
Python
"""桌面客户端入口 — uvicorn 后台服务 + pywebview 桌面窗口。
|
||
|
||
运行方式:
|
||
开发模式: python -m app.desktop (需 pip install pywebview)
|
||
打包后: 双击可执行文件即可
|
||
|
||
职责:
|
||
1. 单实例锁 — 已运行则聚焦已有窗口并退出
|
||
2. 选可用端口 — 从 settings.port 起, 被占则递增
|
||
3. 后台线程起 uvicorn (仅监听 127.0.0.1, 不暴露外网)
|
||
4. 主线程起 pywebview 窗口渲染前端
|
||
5. 窗口关闭 → 优雅停止 uvicorn → 进程退出
|
||
|
||
不含: 业务逻辑、配置持久化、监控告警 (全在 app.main 里)。
|
||
"""
|
||
from __future__ import annotations
|
||
|
||
import errno
|
||
import html
|
||
import logging
|
||
import os
|
||
import socket
|
||
import sys
|
||
import threading
|
||
import time
|
||
import traceback
|
||
from typing import BinaryIO
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
_APP_NAME = "TSP"
|
||
_BASE_PORT = 3018
|
||
_PORT_PROBE_RANGE = 50 # 从 3018 起最多试 50 个端口
|
||
_instance_lock: BinaryIO | None = None
|
||
|
||
|
||
def _failed_html(reason: str | None = None) -> str:
|
||
"""启动失败/超时的整页提示 (后端没起来时窗口也要能说明白)。
|
||
|
||
reason=None → 沿用「启动超时」原文案; 传 reason → 换成具体失败原因,
|
||
文案里的换行按 <br> 渲染, 用户照指引操作即可自助解决。
|
||
"""
|
||
if reason is None:
|
||
title = "TSP 启动超时"
|
||
body = "<p>后端未在预期时间内就绪，请关闭窗口后重新打开。</p>"
|
||
else:
|
||
title = "TSP 启动失败"
|
||
body = f"<p>{html.escape(reason).replace(chr(10), '<br>')}</p>"
|
||
return f"""<!doctype html><html><head><meta charset="utf-8">
|
||
<style>
|
||
html,body{{margin:0;height:100%;background:#0d1117;color:#c9d1d9;
|
||
font-family:"Microsoft YaHei",system-ui,sans-serif;
|
||
display:flex;align-items:center;justify-content:center}}
|
||
.box{{text-align:center;max-width:460px}}
|
||
h1{{font-size:18px;font-weight:500;margin:0 0 12px;color:#f85149}}
|
||
p{{font-size:13px;color:#8b949e;margin:0 0 6px;line-height:1.7}}
|
||
</style></head><body><div class="box">
|
||
<h1>{title}</h1>
|
||
{body}
|
||
<p>反复失败时查看日志：data\\logs\\desktop.log</p>
|
||
</div></body></html>"""
|
||
|
||
|
||
def _ensure_data_dir_writable() -> None:
|
||
"""确保用户数据目录可写 (lifespan 会创建子目录, 这里只验证根目录)。
|
||
|
||
data_dir 在 frozen 模式下指向用户目录 (见 config.py), 非可写会导致
|
||
DuckDB 视图 / parquet 落盘全失败。提前失败胜过启动后乱报错。
|
||
"""
|
||
from app.config import settings
|
||
|
||
data_root = settings.data_dir
|
||
try:
|
||
data_root.mkdir(parents=True, exist_ok=True)
|
||
probe = data_root / ".write_probe"
|
||
probe.write_text("ok", encoding="utf-8")
|
||
probe.unlink(missing_ok=True)
|
||
except Exception as e: # noqa: BLE001
|
||
logger.error("数据目录不可写, 桌面版无法运行: %s (%s)", data_root, e)
|
||
raise
|
||
|
||
|
||
def _precheck_mining_lock() -> str | None:
|
||
"""启动前预检挖掘锁。返回 None = 无冲突; 返回 str = 冲突原因 (仅日志)。
|
||
|
||
app.main 的 lifespan 会抢 data_dir/.mining_process_lock; 被别的后端
|
||
实例 (panel 模式、隧道后端或残留进程) 占住时, 现在 lifespan 只降级
|
||
(挖掘入口 409, 面板可用, 挖掘页提供强制释放), 这里提前留一条日志
|
||
方便排查。试抢成功立即释放, 真正的持有仍由 lifespan 负责。
|
||
"""
|
||
from app.config import settings
|
||
from app.services.mining_process_lock import (
|
||
MiningProcessLock,
|
||
MiningProcessLockError,
|
||
)
|
||
|
||
lock = MiningProcessLock(settings.data_dir)
|
||
try:
|
||
lock.acquire()
|
||
except MiningProcessLockError:
|
||
return (
|
||
"检测到另一个后端实例正在占用数据目录\n"
|
||
"(panel 模式、云隧道或残留进程)\n"
|
||
"请先停止它再打开桌面版：\n"
|
||
"1. 在任务管理器结束残留的 python / uvicorn 进程\n"
|
||
"2. 或重启电脑\n"
|
||
"完成后重新打开本应用"
|
||
)
|
||
lock.release()
|
||
return None
|
||
|
||
|
||
def _acquire_single_instance() -> bool:
|
||
"""单实例锁。已运行返回 False (本进程应退出), 否则 True。
|
||
|
||
文件内容仅供诊断, 实际以系统文件锁为准。进程崩溃后系统会自动释放,
|
||
避免残留 PID 被其他进程复用时永远误判为「已运行」。
|
||
"""
|
||
from app.config import settings
|
||
|
||
global _instance_lock
|
||
if _instance_lock is not None:
|
||
return False
|
||
|
||
lock_path = settings.data_dir / ".desktop.lock"
|
||
stream = lock_path.open("a+b")
|
||
try:
|
||
os.set_inheritable(stream.fileno(), False)
|
||
try:
|
||
if os.name == "nt":
|
||
import msvcrt
|
||
|
||
stream.seek(0)
|
||
msvcrt.locking(stream.fileno(), msvcrt.LK_NBLCK, 1)
|
||
else:
|
||
import fcntl
|
||
|
||
fcntl.flock(stream.fileno(), fcntl.LOCK_EX | fcntl.LOCK_NB)
|
||
except OSError as exc:
|
||
if exc.errno not in (errno.EACCES, errno.EAGAIN, errno.EDEADLK):
|
||
raise
|
||
logger.warning("检测到已有桌面实例持锁, 本进程退出")
|
||
return False
|
||
|
||
stream.seek(0)
|
||
stream.truncate()
|
||
stream.write(str(_current_pid()).encode("ascii"))
|
||
stream.flush()
|
||
_instance_lock = stream
|
||
return True
|
||
finally:
|
||
if _instance_lock is not stream:
|
||
stream.close()
|
||
|
||
|
||
def _release_single_instance() -> None:
|
||
global _instance_lock
|
||
stream, _instance_lock = _instance_lock, None
|
||
if stream is not None:
|
||
# 不 unlink: 删除再重建会让并发启动的进程锁住不同文件, 同时通过检查。
|
||
stream.close()
|
||
|
||
|
||
def _guard_streams() -> None:
|
||
"""windowed 模式 (console=False) 下守护 stdout/stderr。
|
||
|
||
PyInstaller console=False 用 runw.exe 启动器, 不分配控制台, 此时 sys.stdout /
|
||
sys.stderr 可能为 None (或底层句柄无效)。后果:
|
||
- app/__init__.py 早期 reconfigure 遇 None 虽有 hasattr 保护, 但若是个「写入即崩」
|
||
的伪 stream 对象, reconfigure 会成功、后续写却崩;
|
||
- logging.basicConfig() 默认建 StreamHandler(sys.stderr), stderr 为 None 时
|
||
首次写日志调 None.write() 抛 AttributeError, 此时往往在导入早期,
|
||
Python 异常处理未就绪 → 进程直接闪退, try/except 都拦不住。
|
||
|
||
修法: console=False 下把 stdout/stderr 换成丢弃写入的空对象 (devnull),
|
||
让 logging / reconfigure / 任何 print 都安全落地。console=True 不动 (有真控制台)。
|
||
"""
|
||
import os
|
||
|
||
class _NullStream:
|
||
"""丢弃所有写入的空流 (替代 None 的 stdout/stderr)。"""
|
||
def write(self, _s): return 0
|
||
def flush(self): pass
|
||
def reconfigure(self, *a, **kw): pass
|
||
def isatty(self): return False
|
||
def fileno(self): raise OSError("no fileno")
|
||
|
||
# 仅在 stdout/stderr 缺失或不可写时替换 (有真控制台时保持原样)
|
||
for name in ("stdout", "stderr"):
|
||
stream = getattr(sys, name, None)
|
||
if stream is None:
|
||
setattr(sys, name, _NullStream())
|
||
|
||
|
||
def _setup_logging() -> None:
|
||
"""配置日志落盘到 data/logs/desktop.log (10MB×5 切割)。
|
||
|
||
背景: spec 里 console=False (桌面应用不弹黑窗), 导致 logging 默认输出的
|
||
stderr 被吞掉 —— 启动期任何异常用户都看不到, 表现为「双击一闪退出、查无日志」。
|
||
这里经共享的 attach_rotating_log 追加轮转文件 handler, 事后可查; 桌面版
|
||
长期挂机, 不切割的追加日志会无限增长 (用户反馈的磁盘隐患)。
|
||
旧版的 data/desktop.log 由 attach_rotating_log 一次性迁入 data/logs/。
|
||
"""
|
||
from app.logfile import attach_rotating_log
|
||
|
||
attach_rotating_log("desktop.log", max_mb=10, backup_count=5)
|
||
|
||
|
||
def _show_crash(title: str, text: str) -> None:
|
||
"""崩溃时弹原生 MessageBox 提示用户 (仅 Windows)。
|
||
|
||
console=False 下用户看不到任何输出, 崩溃时弹一个原生错误框, 让用户至少
|
||
知道「程序崩了 + 原因」, 并可截图反馈。非 Windows 用日志降级, 不调 ctypes。
|
||
|
||
ctypes 是 Python 标准库, 不引入新依赖; MessageBoxW 是 Unicode 版本 (W 后缀),
|
||
支持中文标题/正文。0x10 = MB_ICONERROR (红色错误图标)。
|
||
"""
|
||
if sys.platform == "win32":
|
||
try:
|
||
import ctypes
|
||
|
||
ctypes.windll.user32.MessageBoxW(0, text, title, 0x10)
|
||
except Exception as e: # noqa: BLE001
|
||
logger.error("弹框失败 (已写日志文件): %s", e)
|
||
else:
|
||
logger.error("%s: %s", title, text)
|
||
|
||
|
||
def _pid_alive(pid: int) -> bool:
|
||
"""检查指定 PID 的进程是否存活。"""
|
||
import os
|
||
|
||
if os.name == "nt":
|
||
# Windows 的 os.kill 没有 POSIX 探测语义: 除 CTRL_C/CTRL_BREAK 外
|
||
# 任何信号值 (含 0) 都是 TerminateProcess 无条件杀进程。存活探测
|
||
# 必须用 OpenProcess + GetExitCodeProcess, 否则单实例检查会误杀
|
||
# 正在运行的实例 (例如探测子进程重启自身时)。
|
||
import ctypes
|
||
|
||
STILL_ACTIVE = 259
|
||
PROCESS_QUERY_LIMITED_INFORMATION = 0x1000
|
||
kernel32 = ctypes.windll.kernel32 # type: ignore[attr-defined]
|
||
handle = kernel32.OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, False, pid)
|
||
if not handle:
|
||
return False
|
||
try:
|
||
exit_code = ctypes.c_ulong()
|
||
if not kernel32.GetExitCodeProcess(handle, ctypes.byref(exit_code)):
|
||
return False
|
||
return exit_code.value == STILL_ACTIVE
|
||
finally:
|
||
kernel32.CloseHandle(handle)
|
||
else:
|
||
try:
|
||
os.kill(pid, 0) # signal 0 = 探测存活, 不实际发信号
|
||
return True
|
||
except OSError:
|
||
return False
|
||
|
||
|
||
def _current_pid() -> int:
|
||
import os
|
||
|
||
return os.getpid()
|
||
|
||
|
||
def _find_free_port(start: int, count: int = _PORT_PROBE_RANGE) -> int:
|
||
"""从 start 起找第一个可用端口。全部被占则返回 start (交给 uvicorn 报错)。"""
|
||
for port in range(start, start + count):
|
||
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
|
||
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
|
||
try:
|
||
s.bind(("127.0.0.1", port))
|
||
return port
|
||
except OSError:
|
||
continue
|
||
return start
|
||
|
||
|
||
def _run_uvmicorn(port: int, ready_event: threading.Event) -> None:
|
||
"""后台线程: 启动 uvicorn 服务。ready_event 在线程退出时置位 (通知主线程)。"""
|
||
import uvicorn
|
||
|
||
try:
|
||
# 延迟 import app, 确保配置层已就绪 (frozen 检测在 config.py 导入时完成)
|
||
# 放进 try: app.main 模块导入 (含 app.api.* 一长串 import) 若失败,
|
||
# 异常必须落到 except 记录, 否则主线程只看到「后端超时」而查无 traceback。
|
||
from app.main import app
|
||
except Exception:
|
||
logger.exception("后端模块导入失败 (app.main 或其依赖)")
|
||
ready_event.set()
|
||
return
|
||
|
||
config = uvicorn.Config(
|
||
app,
|
||
host="127.0.0.1", # 仅本机, 不暴露外网 (桌面版无需远程访问)
|
||
port=port,
|
||
log_level="info",
|
||
access_log=False, # 桌面版不需要访问日志
|
||
loop="auto",
|
||
# log_config=None: uvicorn 默认给 uvicorn.* 挂独立 console handler 且
|
||
# propagate=False, windowed 模式下 stderr 已被守护替换, lifespan 启动
|
||
# 失败的 "Application startup failed" 永远到不了 desktop.log (失败页
|
||
# 承诺的排查入口)。置 None 让 uvicorn 日志走 root 的 FileHandler 落盘。
|
||
log_config=None,
|
||
)
|
||
server = uvicorn.Server(config)
|
||
|
||
# 线程结束时通知主线程 (无论正常退出还是异常)
|
||
def _signal_done(*exc):
|
||
ready_event.set()
|
||
server.config.callback_notify = None # 不用 notify 机制
|
||
|
||
try:
|
||
server.run()
|
||
except Exception:
|
||
# server.run() 内部跑 lifespan 启动链, 任一步抛异常都会冒到这里。
|
||
# 不捕获则线程静默死亡, 主线程 _wait_for_server 傻等满 60s 后报「超时」,
|
||
# 真正的崩溃原因 (如某个原生库加载失败 / 缺 hidden import) 永远看不到。
|
||
logger.exception("uvicorn 后端启动/运行失败")
|
||
finally:
|
||
ready_event.set()
|
||
|
||
|
||
def _wait_for_server(
|
||
port: int, timeout: float = 60.0, done_event: threading.Event | None = None
|
||
) -> bool:
|
||
"""轮询 health 接口直到后端就绪或超时。
|
||
|
||
比 monkey-patch uvicorn 内部方法更健壮, 不依赖版本内部实现。
|
||
|
||
done_event 置位表示后端线程已退出 (lifespan 失败被 uvicorn 吞掉时不向
|
||
调用方抛异常, 只表现为 server.run() 正常返回) —— 此时不可能出现就绪,
|
||
立即返回 False, 不再傻等满 timeout 把「启动失败」伪装成「超时」。
|
||
"""
|
||
import urllib.request
|
||
import urllib.error
|
||
|
||
url = f"http://127.0.0.1:{port}/health"
|
||
deadline = time.monotonic() + timeout
|
||
while time.monotonic() < deadline:
|
||
if done_event is not None and done_event.is_set():
|
||
return False
|
||
try:
|
||
with urllib.request.urlopen(url, timeout=2) as r:
|
||
if r.status == 200:
|
||
return True
|
||
except (urllib.error.URLError, ConnectionError, OSError):
|
||
pass
|
||
time.sleep(0.5)
|
||
return False
|
||
|
||
|
||
def _open_window(url: str) -> None:
|
||
"""主线程: 用 pywebview 打开桌面窗口。"""
|
||
import webview # type: ignore[import-not-found]
|
||
|
||
# pywebview 5.0 起默认取消所有文件下载 (Edge WebView2 / WKWebView 一致):
|
||
# 配置备份导出、回测 CSV 等 <a download> 直链会点击无反应, 必须在 start() 前显式开启。
|
||
webview.settings["ALLOW_DOWNLOADS"] = True
|
||
window = webview.create_window(
|
||
_APP_NAME,
|
||
url,
|
||
width=1440,
|
||
height=900,
|
||
min_size=(1024, 700),
|
||
# 桌面版固定单窗口, 禁用外部浏览器跳转
|
||
confirm_close=False,
|
||
)
|
||
# 继承隐藏启动标志时首次显示可能被抑制, GUI 就绪后再显式显示 (同失败页)。
|
||
webview.start(lambda: window.show(), debug=False)
|
||
|
||
|
||
def _open_static_window(html_text: str) -> None:
|
||
"""只弹一个静态内容窗口 (预检失败等场景: 后端根本没起, 无需等待切换)。"""
|
||
import webview # type: ignore[import-not-found]
|
||
|
||
window = webview.create_window(
|
||
_APP_NAME,
|
||
html=html_text,
|
||
width=560,
|
||
height=360,
|
||
confirm_close=False,
|
||
)
|
||
# 启动器以 SW_HIDE 隐藏控制台, Windows 会把隐藏标志传给进程第一个窗口
|
||
# (即本窗口), 故 GUI 循环就绪后显式 show (同 _open_window 的就绪处理)
|
||
webview.start(lambda: window.show(), debug=False)
|
||
|
||
|
||
def main() -> int:
|
||
"""桌面客户端主入口。返回进程退出码。"""
|
||
# 必须最先执行: console=False 下 stdout/stderr 可能无效, 不守护会导致
|
||
# 后续 logging.basicConfig 创建的 StreamHandler 写日志时进程崩溃。
|
||
_guard_streams()
|
||
logging.basicConfig(
|
||
level=logging.INFO,
|
||
format="%(asctime)s [%(levelname)s] %(name)s: %(message)s",
|
||
)
|
||
# 追加文件日志: console=False 下 stderr 被吞, 必须落盘否则查无对证。
|
||
# 放在 basicConfig 之后 (它先建好 root logger 的格式), 这里只追加 handler。
|
||
_setup_logging()
|
||
|
||
# TdxQuant 探测 worker 哨兵 (见 probe.run): 打包版以「重执行自身 + 哨兵」
|
||
# 拉起独立探测进程, 必须先于单实例锁/后端/窗口等任何桌面启动动作分发。
|
||
if os.environ.get("TDX_PROBE_WORKER") == "1":
|
||
from app.plugins.tdxquant.probe import run_worker
|
||
|
||
return run_worker()
|
||
|
||
instance_acquired = False
|
||
try:
|
||
_ensure_data_dir_writable()
|
||
|
||
if not _acquire_single_instance():
|
||
return 0
|
||
instance_acquired = True
|
||
|
||
# 撞挖掘锁只降级, 其他预检异常必须同样落盘、弹框并释放桌面锁。
|
||
conflict = _precheck_mining_lock()
|
||
if conflict is not None:
|
||
logger.warning("启动预检: %s", conflict.replace("\n", " "))
|
||
|
||
port = _find_free_port(_BASE_PORT)
|
||
logger.info("桌面版后端将监听 127.0.0.1:%d", port)
|
||
|
||
# 后台线程起 uvicorn
|
||
ready = threading.Event()
|
||
server_thread = threading.Thread(
|
||
target=_run_uvmicorn, args=(port, ready), daemon=True,
|
||
name="uvicorn",
|
||
)
|
||
server_thread.start()
|
||
|
||
# 轮询 health 接口等后端就绪 (含 lifespan 初始化, 最多 60s);
|
||
# ready 由后端线程 finally 置位 → 线程已退出时立刻返回, 不等满超时
|
||
if not _wait_for_server(port, timeout=60.0, done_event=ready):
|
||
reason = (
|
||
"后端进程已提前退出，详情见 data\\logs\\desktop.log"
|
||
if ready.is_set() else None
|
||
)
|
||
logger.error("后端未就绪 (reason=%s), 桌面版退出", reason)
|
||
_open_static_window(_failed_html(reason))
|
||
return 1
|
||
|
||
url = f"http://127.0.0.1:{port}"
|
||
logger.info("打开桌面窗口: %s", url)
|
||
_open_window(url)
|
||
|
||
# 窗口关闭后, 进程退出 (daemon 线程会被回收)
|
||
logger.info("窗口已关闭, 桌面版退出")
|
||
return 0
|
||
except KeyboardInterrupt:
|
||
return 0
|
||
except Exception:
|
||
# 顶层兜底: console=False 下未捕获异常会「一闪退出」且无任何反馈。
|
||
# 写完整 traceback 到 data/desktop.log, 并弹原生 MessageBox 让用户截图反馈。
|
||
# 必须排在 KeyboardInterrupt 之后 —— Exception 是基类, 在前会遮蔽它。
|
||
logger.exception("桌面客户端启动失败")
|
||
_show_crash("TSP 启动失败", traceback.format_exc())
|
||
return 1
|
||
finally:
|
||
if instance_acquired:
|
||
_release_single_instance()
|
||
|
||
|
||
if __name__ == "__main__":
|
||
import multiprocessing
|
||
|
||
multiprocessing.freeze_support()
|
||
sys.exit(main())
|