C++程序Segmentation Fault问题：Valgrind检测到Invalid write of size 8

开发求解二维热方程的C++程序时出现Segmentation Fault，怀疑是内存访问问题但无法定位具体原因，相关代码及错误信息如下：

代码片段 #

myclass.h #

#include <vector>
using vector_double = std::vector<double>;
using matrix2d = std::vector<std::vector<double>>;

class Myclass{
    protected:
        matrix2d matrix_;
        vector_double vector_;
        size_t dim;
        size_t maxT;
        double initValue;

    public:
        Myclass();
        void routine();
};

myclass.cpp #

#include "myclass.h"

Myclass::Myclass()
{   
    size_t dim = 1001;
    size_t maxT = 1001;
    double initValue = 13.0;
    matrix2d matrix_(maxT+1, vector_double(dim,initValue));
    vector_double vector_(dim);
    
}

void Myclass::routine()
{
    vector_[0] = initValue;

    for(size_t t=1; t<maxT; t++)
    {
        vector_[0] = matrix_[t][0] + matrix_[t][0];
        for(size_t i=1; i < dim-1; i++)
        {
            vector_[i] = matrix_[t][i-1] + matrix_[t][i] + matrix_[t][i+1];
        }
    }
}

so.cpp #

#include "myclass.h"

int main()
{
    Myclass myobject;
    myobject.routine();

return 0;
}

编译与运行命令 #

g++ -g -O1 *.cpp -o out
valgrind --leak-check=yes ./out

错误信息 #

==25934== Memcheck, a memory error detector
==25934== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==25934== Using Valgrind-3.15.0 and LibVEX; rerun with -h for copyright info
==25934== Command: ./out
==25934== 
==25934== Invalid write of size 8
==25934==    at 0x109217: Myclass::routine() (myclass.cpp:15)
==25934==    by 0x109577: main (so.cpp:7)
==25934==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==25934== 
==25934== 
==25934== Process terminating with default action of signal 11 (SIGSEGV)
==25934==  Access not within mapped region at address 0x0
==25934==    at 0x109217: Myclass::routine() (myclass.cpp:15)
==25934==    by 0x109577: main (so.cpp:7)
==25934==  If you believe this happened as a result of a stack
==25934==  overflow in your program's main thread (unlikely but
==25934==  possible), you can try to increase the size of the
==25934==  main thread stack using the --main-stacksize= flag.
==25934==  The main thread stack size used in this run was 8388608.
==25934== 
==25934== HEAP SUMMARY:
==25934==     in use at exit: 0 bytes in 0 blocks
==25934==   total heap usage: 1,006 allocs, 1,006 frees, 8,136,784 bytes allocated
==25934== 
==25934== All heap blocks were freed -- no leaks are possible
==25934== 
==25934== For lists of detected and suppressed errors, rerun with: -s
==25934== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
Segmentation fault (core dumped)

问题根源是构造函数内的局部变量遮蔽了类成员变量：你在构造函数中定义了同名的局部变量dim、maxT、initValue、matrix_、vector_，导致类的成员变量始终处于未初始化状态（dim、maxT为随机值，matrix_、vector_是空容器），在routine()中访问这些未初始化的成员变量就会触发内存错误。

修正后的构造函数（直接赋值成员变量） #

Myclass::Myclass()
{   
    dim = 1001;
    maxT = 1001;
    initValue = 13.0;
    matrix_ = matrix2d(maxT+1, vector_double(dim,initValue));
    vector_ = vector_double(dim);
}

更推荐的写法（成员初始化列表，效率更高） #

Myclass::Myclass() 
    : dim(1001), maxT(1001), initValue(13.0),
      matrix_(maxT+1, vector_double(dim, initValue)),
      vector_(dim)
{}

另外，routine()中的循环逻辑本身是合法的：

• t<maxT循环中，matrix_[t]的索引范围是1到1000，对应matrix_的大小1002（索引0到1001），无越界
• i < dim-1循环中，i最大为999，i+1=1000对应vector_和matrix_[t]的合法索引

修正后重新编译运行，内存错误即可解决。

内容的提问来源于stack exchange，提问作者nico_so