c++ - 应用程序因 _Unwind_Resume 中止错误而崩溃

Question

调用堆栈:

#0 0x00007faf7fdb8ed5 in raise () from /lib/libc.so.6
#1 0x00007faf7fdba3f3 in abort () from /lib/libc.so.6
#2 0x00007faf8063c294 in _gnu_cxx::_verbose_terminate_handler () from /usr/lib/libstdc++.so.6
#3 0x00007faf8063a696 in ?? () from /usr/lib/libstdc++.so.6
#4 0x00007faf8063988b in ?? () from /usr/lib/libstdc++.so.6
#5 0x00007faf8063a458 in _gxx_personality_v0 () from /usr/lib/libstdc++.so.6
#6 0x00007faf800eacb3 in ?? () from /lib/libgcc_s.so.1
#7 0x00007faf800ead78 in _Unwind_Resume () from /lib/libgcc_s.so.1
#8 0x0000000000a0ea6c in ~ServletRequest (this=0x7faf60a156c0) at ../myapp/servlets/server.cpp:124
#9 0x00000000009d8be2 in boost::detail::sp_counted_impl_p<MyApp::ServletRequest>::dispose (this=<value optimized out> at /usr/include/boost/checked_delete.hpp:34
#10 0x00000000006f5569 in ~shared_count (this=<value optimized out> at /usr/include/boost/smart_ptr/detail/sp_counted_base_gcc_x86.hpp:145
#11 0x00000000009d4a59 in MyApp::Dispatcher::request (this=0x19413b8, req={px = 0x7faf732dfc70, pn = {pi = 0x7faf732dfb60}}) at /usr/include/boost/smart_ptr/shared_ptr.hpp:169
#12 0x00000000009afd9e in boost::detail::function::void_function_ref_invoker1<MyApp::Dispatcher, void, boost::shared_ptr<MyLib::HTTP::ServerRequest> >::invoke (function_obj_ptr=<value optimized out>, a0=<value optimized out> at ../libtriton/mordor/mordor/http/servlet.h:28
#13 0x0000000000cd2bb3 in boost::function1<void, boost::shared_ptr<MyLib::HTTP::ServerRequest> >::operator() (this=<value optimized out>, a0=<value optimized out> at /usr/include/boost/function/function_template.hpp:1013

任何人都遇到过这个问题，可以阐明为什么在 _Unwind_Resume 中发生中止吗？

Answer 1

写了一个简单的测试来模拟这个问题:

class A {
public:
    typedef boost::shared_ptr<A> ptr;
    A(std::string name = "default") : m_name(name) { }
    void throwInDestor() {
        if (m_name == "throwInDestructor")
            throw 34;
    }
    ~A(){
        throwInDestor();
    };
    void throwName() {
        throw m_name;
    }

    std::string m_name;
};

void callfunc(A::ptr a) {
    a->throwName();
}

void callf1(A::ptr a) {
    A::ptr b(new A("throwInDestructor"));
    callfunc(a);
}

void callf2() {
    A::ptr a(new A("call2"));
    callf1(a);
}

int main() {
    try {
        try {
            callf2();
        } catch (...) {
            std::cout << "inner: " << boost::diagnostic_information(boost::current_exception()) << std::endl;
        }
    } catch (...) {
        std::cout << "outer: " << boost::diagnostic_information(boost::current_exception()) << std::endl;
    }
}

用g++编译，成功重现_Unwind_Resume错误。_Unwind_Resume 错误可以通过在 class A 的析构函数中捕获异常来忽略，请参见以下内容:

~A(){
    try {
        throwInDestor();
    } catch (...) {
        std::cout << "A::~A: " << boost::diagnostic_information(boost::current_exception()) << std::endl;
    }
};

std::string 异常被抛出，并且在主函数中找到了 catch 处理程序。然后在 callf1 中进行清理工作时发生崩溃，而 std::string 异常正在传播到 main因为第一个异常是在展开和清理工作的过程中，但是int 异常是在函数callf1 的A 的析构函数中抛出的。因此解决方案可以是在 A 的析构函数中立即捕获 int 异常，如上所述。

我不知道它为什么有效，但它确实有效。