[Bug] 升级后 GUI 闪退：runtime handoff 因 PID 复用陷入死循环

[Deepfriedmonkey-9527]

Summary / 问题概述

从 0.2.37 升级到 0.3.7 后，Kun GUI 无法正常启动。打开后短暂显示"有后台存在"，随即窗口消失、进程退出。重装软件无法修复。

根因是 ~/.kun/data/runtime.json 中记录的旧运行时 PID 已被 Windows 系统回收并分配给无关进程，而 update-handoff 模块在停止旧运行时前不校验 PID 归属，导致 handoff 必然失败且无降级路径，形成死循环。

Steps to reproduce / 复现步骤

1.安装 Kun 0.2.37，正常使用，运行时进程 A 监听在某端口（如 PID 15016, port 25062）
2.Kun 自动下载 0.3.7 更新并触发安装，新版本文件部署到安装目录，build ID 从 dc2ddd696ec8 变为 4d51719d6435
3.安装过程中或安装后，旧运行时进程 A 退出（可能因安装器替换文件导致），但 ~/.kun/data/runtime.json 仍记录着 PID 15016 的连接信息
4.Windows 将 PID 15016 回收并分配给系统服务 WMIRegistrationService.exe
5.用户启动 Kun GUI
此时 GUI 进入 handoff 流程：检测到磁盘 build（4d51719d6435）与 runtime.json 中的 build（dc2ddd696ec8）不一致，尝试停止旧运行时（PID 15016 / port 25062）。由于该 PID 已不属于 Kun，停止请求失败，GUI 直接退出。

Expected behavior / 期望行为

1.update-handoff 在 stop-runtimes 阶段应校验目标 PID 是否仍属于 Kun 运行时（如检查进程可执行文件路径、或向 port 发 /health 请求校验 instanceId）
2.若 PID 已不属于 Kun（进程不存在、或进程非 Kun），应视为旧运行时已自然退出，清除 runtime.json 后继续启动新版本
3.handoff 失败后应有降级路径，而非直接放弃启动——至少允许以当前磁盘 build 启动
4.gui-version-state.json 中的 pendingUpdate 在 handoff 失败时不应永久保留，避免每次启动都触发同样的失败流程
5.~/.kun/.data.kun-runtime-writer-claims/ 目录中存在非法文件名时，应跳过而非阻断启动

Actual behavior / 实际行为

1.GUI 向已被系统服务占用的 PID 15016 发送停止请求，收到 runtime_stop_failed，然后直接退出，不进入主界面
2.每次启动都重复同样的 handoff → stop-runtimes → failed → 退出循环，pendingUpdate 始终不被清除，升级永远无法完成
3.重装软件不能修复——因为 runtime.json 和 writer-claims 位于用户数据目录 ~/.kun/data/，不在安装目录内

Kun version / Kun 版本

0.3.7

Operating system / 操作系统

Windows 11 Insider Preview 26340.9233

Logs / 日志

1. GUI 日志：首次出现 handoff 失败（kun-2026-09-02.log）
9 月 2 日 15:43 UTC（北京时间 23:43）首次出现 installed-build-change + runtime_stop_failed：


[2026-09-02T15:43:32.400Z] [INFO] [startup] Desktop startup entered. - detail: {
  "platform": "win32",
  "packaged": true
}
[2026-09-02T15:43:32.599Z] [INFO] [update-handoff] Kun owner handoff discover - detail: {
  "reason": "installed-build-change",
  "phase": "discover",
  "elapsedMs": 0,
  "targetBuildId": "4d51719d6435"
}
[2026-09-02T15:43:32.601Z] [INFO] [update-handoff] Kun owner handoff discover - detail: {
  "reason": "installed-build-change",
  "phase": "discover",
  "elapsedMs": 2,
  "targetBuildId": "4d51719d6435",
  "probeClassification": "runtime-discovery-compatible"
}
[2026-09-02T15:43:32.601Z] [INFO] [update-handoff] Kun owner handoff quiesce-runtimes - detail: {
  "reason": "installed-build-change",
  "phase": "quiesce-runtimes",
  "elapsedMs": 2,
  "targetBuildId": "4d51719d6435"
}
[2026-09-02T15:43:33.119Z] [WARN] [update-handoff] Kun owner handoff stop-runtimes: failed — detail: {
  "reason": "installed-build-change",
  "phase": "stop-runtimes",
  "elapsedMs": 520,
  "targetBuildId": "4d51719d6435",
  "result": "failed",
  "code": "runtime_stop_failed",
  "ownerKind": "runtime",
  "flavor": "production",
  "pid": 15016,
  "instanceId": "e50571ea-0373-41fe-bd1e-1d626ac1fb82",
  "port": 25062,
  "buildId": "dc2ddd696ec83cf05bb7d420df2987e99fea42f97c6f4ab07153bc53ff8cd24c"
}
此日志之后无任何其他内容——GUI 直接退出，未进入主界面。

2. GUI 日志：次日反复启动失败（kun-2026-09-03.log）
9 月 3 日用户多次尝试启动，每次都重复同样的 handoff → stop-runtimes → failed 循环，共 7 次，间隔从几秒到几分钟不等：


[2026-09-03T12:22:58.647Z] [INFO] [startup] Desktop startup entered. - detail: {
  "platform": "win32",
  "packaged": true
}
[2026-09-03T12:22:58.877Z] [INFO] [update-handoff] Kun owner handoff discover - detail: {
  "reason": "installed-build-change",
  "phase": "discover",
  "elapsedMs": 0,
  "targetBuildId": "4d51719d6435"
}
[2026-09-03T12:22:58.880Z] [INFO] [update-handoff] Kun owner handoff discover - detail: {
  "reason": "installed-build-change",
  "phase": "discover",
  "elapsedMs": 3,
  "targetBuildId": "4d51719d6435",
  "probeClassification": "runtime-discovery-compatible"
}
[2026-09-03T12:22:58.880Z] [INFO] [update-handoff] Kun owner handoff quiesce-runtimes - detail: {
  "reason": "installed-build-change",
  "phase": "quiesce-runtimes",
  "elapsedMs": 3,
  "targetBuildId": "4d51719d6435"
}
[2026-09-03T12:22:59.524Z] [WARN] [update-handoff] Kun owner handoff stop-runtimes: failed — detail: {
  "reason": "installed-build-change",
  "phase": "stop-runtimes",
  "elapsedMs": 647,
  "targetBuildId": "4d51719d6435",
  "result": "failed",
  "code": "runtime_stop_failed",
  "ownerKind": "runtime",
  "flavor": "production",
  "pid": 15016,
  "instanceId": "e50571ea-0373-41fe-bd1e-1d626ac1fb82",
  "port": 25062,
  "buildId": "dc2ddd696ec83cf05bb7d420df2987e99fea42f97c6f4ab07153bc53ff8cd24c"
}

[2026-09-03T12:25:08.403Z] [INFO] [startup] Desktop startup entered.    ← 第 2 次尝试
[2026-09-03T12:25:08.993Z] [WARN] [update-handoff] stop-runtimes: failed — pid: 15016, code: "runtime_stop_failed"

[2026-09-03T12:25:19.102Z] [INFO] [startup] Desktop startup entered.    ← 第 3 次尝试
[2026-09-03T12:25:19.693Z] [WARN] [update-handoff] stop-runtimes: failed — pid: 15016, code: "runtime_stop_failed"

[2026-09-03T12:25:32.842Z] [INFO] [startup] Desktop startup entered.    ← 第 4 次尝试
[2026-09-03T12:25:33.432Z] [WARN] [update-handoff] stop-runtimes: failed — pid: 15016, code: "runtime_stop_failed"

[2026-09-03T12:25:40.523Z] [INFO] [startup] Desktop startup entered.    ← 第 5 次尝试
[2026-09-03T12:25:41.132Z] [WARN] [update-handoff] stop-runtimes: failed — pid: 15016, code: "runtime_stop_failed"

[2026-09-03T12:32:35.041Z] [INFO] [startup] Desktop startup entered.    ← 第 6 次尝试
[2026-09-03T12:32:35.893Z] [WARN] [update-handoff] stop-runtimes: failed — pid: 15016, code: "runtime_stop_failed"

[2026-09-03T12:33:06.110Z] [INFO] [startup] Desktop startup entered.    ← 第 7 次尝试
[2026-09-03T12:33:06.709Z] [WARN] [update-handoff] stop-runtimes: failed — pid: 15016, code: "runtime_stop_failed"
注意：每次 stop-runtimes 的 elapsedMs 在 440-650ms 之间，说明并非超时，而是请求被立即拒绝或返回错误。

3. 过期 runtime.json 内容

{
  "version": 2,
  "instanceId": "e50571ea-0373-41fe-bd1e-1d626ac1fb82",
  "pid": 15016,
  "startedAt": "2026-09-02T13:12:36.129Z",
  "host": "127.0.0.1",
  "port": 25062,
  "baseUrl": "http://127.0.0.1:25062",
  "runtimeToken": "7coozwoMWxPPh8fAhga8Y95QCrN2J7q4WtIwjs3Dqno",
  "insecure": false,
  "serviceVersion": "0.2.37",
  "buildId": "dc2ddd696ec83cf05bb7d420df2987e99fea42f97c6f4ab07153bc53ff8cd24c",
  "launchMode": "shared",
  "logPath": "C:\\Users\\cleve\\.kun\\data\\logs\\runtime.log"
}
该文件记录了旧版运行时（版本 0.2.37，PID 15016），但该 PID 此时已不属于 Kun 进程。

4. PID 归属验证

> tasklist | findstr 15016
WMIRegistrationService.ex    15016 Services                   0     16,184 K
PID 15016 已被 Windows 系统服务 WMIRegistrationService.exe 占用，与 Kun 无关。

5. 版本状态（升级未完成）

// %AppData%\Kun\gui-version-state.json
{
  "lastSeenVersion": "0.2.37",
  "pendingUpdate": {
    "version": "0.3.7"
  }
}
pendingUpdate 一直存在，说明 handoff 从未成功完成过，升级状态始终未推进。

6. writer-claims 残留文件

// ~/.kun/.data.kun-runtime-writer-claims/claim-42732-92c84687-8e3f-4e64-94b6-546b1c2d784d.json
{
  "schemaVersion": 1,
  "kind": "runtime",
  "pid": 42732,
  "token": "92c84687-8e3f-4e64-94b6-546b1c2d784d",
  "startedAt": "2026-09-02T13:12:17.728Z",
  "dataDir": "C:\\Users\\cleve\\.kun\\data"
}
PID 42732 同样已不存在。此残留文件在清理时若仅改后缀名（如 .stale）留在目录内，Kun 会因文件名格式校验失败而拒绝启动，报错如下：


[kun-gui] startup failed: Kun Runtime writer claim name is invalid:
  C:\Users\cleve\.kun\.data.kun-runtime-writer-claims\claim-42732-92c84687-8e3f-4e64-94b6-546b1c2d7

Checklist / 检查清单

• I searched existing issues before opening this report. / 我已经搜索过现有 issue。
• I removed sensitive information from logs and screenshots. / 我已经从日志和截图中移除敏感信息。

[XingYu-Zhong]

0.3.8