Conversation
… exit In the shutdown lambda, after deleting dccManager and before deleting app, explicitly clear and wait for the global thread pool to finish. This prevents a race condition in Qt 6.8 where ~QCoreApplication() sets self = nullptr before calling QThreadPool::globalInstance()->waitForDone(), causing worker threads to dereference the null self pointer via notifyInternal2(). PMS: BUG-375601 fix: 退出前清空全局线程池防止崩溃 在 shutdown lambda 中，delete dccManager 之后、delete app 之前，显式调用 QThreadPool::globalInstance()->clear() 和 waitForDone() 清空并等待全局 线程池。防止 Qt 6.8 中 ~QCoreApplication() 先将 self 置 nullptr 再 waitForDone() 时，工作线程通过 notifyInternal2() 解引用空指针导致 SIGSEGV。 PMS: BUG-375601
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: mhduiy The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
Reviewer's guide (collapsed on small PRs)Reviewer's GuideUpdates the managed-memory shutdown sequence to clear and synchronously drain Qt's global thread pool after deleting dccManager but before deleting the application, preventing in-flight image-loading workers from accessing a destroyed QCoreApplication during exit. Sequence diagram for safe application shutdownsequenceDiagram
participant Shutdown as shutdown
participant Manager as dccManager
participant Pool as QThreadPool.globalInstance()
participant App as QCoreApplication
Shutdown->>Manager: delete dccManager
Shutdown->>Pool: clear()
Shutdown->>Pool: waitForDone()
Pool-->>Shutdown: all global tasks completed
Shutdown->>App: delete app
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
deepin pr auto review🤖 AI 代码审查报告总体评分： 📊 总体评价
🔍 详细分析1. 语法逻辑 ✅评价： 优秀 ✅ 通过 潜在问题： 建议： 语法正确，逻辑清晰。新增的 #include 头文件包含正确，QThreadPool::globalInstance()->clear() 和 waitForDone() 调用语法无误。代码放置位置正确：在 delete dccManager 之后、delete app 之前的 shutdown lambda 中，符合修复 Qt 6.8 竞态条件的逻辑要求。clear() 先清除未开始的待处理任务，waitForDone() 再等待正在运行的任务完成，操作顺序合理。 2. 代码质量 ✅评价： 优秀 ✅ 通过 潜在问题：
建议： 建议在 QThreadPool::globalInstance()->clear() 前添加行内注释，简要说明为何需要在 delete app 前清空线程池，例如：// Drain thread pool before app deletion to prevent Qt 6.8 crash (BUG-375601) 3. 代码性能 ✅评价： 优秀 ✅ 通过 潜在问题： 建议： 性能良好，资源使用合理。waitForDone() 是阻塞调用，但在应用退出场景下是正确且必要的行为。clear() 清除未开始的待处理任务，避免不必要的等待。代码仅在 DCC_ENABLE_MEMORY_MANAGEMENT 宏定义时编译，不影响正常运行的性能。 4. 代码安全 🔒评价： 优秀 ✅ 通过
安全漏洞详情： 建议： 存在0个安全漏洞。本次变更不涉及用户输入处理、网络通信、文件操作等安全敏感场景。代码仅调用 Qt 框架的标准线程池管理接口，无注入风险、无敏感信息泄露、无权限绕过风险。修复实际上提升了程序的稳定性，防止了 SIGSEGV 崩溃。 💡 改进建议代码示例// 在 shutdown lambda 中，delete dccManager 之后、delete app 之前
// 显式清空并等待全局线程池完成
// 防止 Qt 6.8 中 ~QCoreApplication() 先置 self=nullptr 再 waitForDone()
// 导致工作线程通过 notifyInternal2() 解引用空指针 (BUG-375601)
QThreadPool::globalInstance()->clear();
QThreadPool::globalInstance()->waitForDone();
delete app;本报告由 AI 代码审查工具自动生成 |
|
TAG Bot New tag: 6.1.109 |
This branch has not been deployed
2 participants
Root Cause Analysis
dde-control-center crashes with SIGSEGV during application exit under reboot stress testing. The root cause is a race condition in Qt 6.8's
~QCoreApplication()destructor: it setsQCoreApplication::self = nullptrbefore callingQThreadPool::globalInstance()->waitForDone(). Meanwhile,DccImageProvideruses the global thread pool for async image loading, and worker threads still running inQThread::exec()callnotifyInternal2()which dereferences the now-nullselfpointer. The shutdown lambda inmain.cppdeletesdccManager(cleaning the local thread pool) but does not drain the global thread pool before deletingapp.Fix
Added
QThreadPool::globalInstance()->clear()andQThreadPool::globalInstance()->waitForDone()in theshutdownlambda, betweendelete dccManageranddelete app. This ensures all global thread pool tasks are cleared and completed whileQCoreApplication::selfis still valid, preventing the destructor race condition. Also added#include <QThreadPool>.Change Safety Assessment
Code Safety
shutdownlambda was recently introduced (2026-09-09) as part of the--listfeature, not a previous bug fix — no risk of reverting a historical fix.#ifdef DCC_ENABLE_MEMORY_MANAGEMENT, only affecting the exit path; no function signatures or public APIs are changed.Business Impact Scope
The change only affects the application exit path (shutdown/reboot/logout scenarios). All normal control center operations (display, sound, network, account settings, etc.) are unaffected. The fix specifically prevents crashes during system reboot stress testing where dde-control-center exits while global thread pool tasks are still in-flight.
Verification Suggestion
Run reboot stress testing (5000+ cycles) to verify no SIGSEGV crashes occur during dde-control-center exit. Verify normal exit via window close and DBus quit still works without delay.
根因分析
dde-control-center 在重启压力测试中退出时发生 SIGSEGV 崩溃。根因是 Qt 6.8 的
~QCoreApplication()析构函数先将QCoreApplication::self置为nullptr，再调用QThreadPool::globalInstance()->waitForDone()。此时DccImageProvider通过全局线程池执行的异步图片加载任务仍在运行，工作线程在QThread::exec()中调用notifyInternal2()解引用已为空的self指针。main.cpp的shutdownlambda 在delete dccManager（清理本地线程池）后直接delete app，未在两者之间清空全局线程池。修复方案
在
shutdownlambda 中delete dccManager与delete app之间新增QThreadPool::globalInstance()->clear()和QThreadPool::globalInstance()->waitForDone()，确保在QCoreApplication::self仍有效时清空并等待全局线程池所有任务完成，消除析构竞态。同时添加#include <QThreadPool>。改动安全评估
代码安全评估
shutdownlambda 为 2026-09-09 新引入的--list功能代码，非历史 bug 修复产物，不存在撤销历史修复的回归风险。#ifdef DCC_ENABLE_MEMORY_MANAGEMENT保护内，仅影响退出路径，不修改任何函数签名或公开接口。业务影响范围
本次修改仅影响应用退出流程（关机/重启/注销场景）。控制中心的各项设置功能（显示、声音、网络、账户等）正常运行不受影响。修复专门针对系统重启压力测试中 dde-control-center 退出时全局线程池任务仍在运行导致的崩溃。
验证建议
执行重启压力测试（5000+ 次循环）验证 dde-control-center 退出时不再发生 SIGSEGV 崩溃。验证通过窗口关闭和 DBus quit 正常退出时无延迟。
Summary by Sourcery
Bug Fixes: