C++：导致访问冲突的向量元素分配

Question

在 C++ 中将字符串分解为单独的行和参数（作为二维向量）时，在尝试解析函数之间的向量时可能会产生有趣的访问冲突问题。在代码示例中，已多次尝试确保传入和传出函数的数据是独立的对象，绝不是引用。

segregate.hpp

#pragma once

#include <vector>
#include <string>

/*
  Purpose:
    To to take a whole file as a string,
    and break it up into individual words
*/

namespace Segregate {
  // Module types
  typedef std::vector< std::string > ParamArray;
  struct StrCommand{
    unsigned long line;
    ParamArray param;
  };
  typedef std::vector< StrCommand > StrCommands;

  bool IsParamBreak(char val);
  bool IsLineBreak(char val);

  ParamArray Parameterize(std::string str);
  StrCommands Fragment(std::string str);
}


#include "./segregate.cpp"

segregate.cpp

#include "./segregate.hpp"

namespace Segregate{
    bool IsParamBreak(char val){
        if (val == ' '){
            return true;
        }else if (val == '\t'){
            return true;
        }

        return false;
    };
    bool IsLineBreak(char val){
        if (val == '\n'){
            return true;
        }

        return false;
    };

    // Splits a single line into individual parameters
    ParamArray Parameterize(std::string str){
        str.append(" "); // Ensures that the loop will cover all segments
        unsigned long length = str.size();
        unsigned long comStart = 0;
        ParamArray res;


        // Ignore carrage returns
        //  Windows artifact
        if (str[0] == '\r'){
            comStart = 1;
        }


        // Ignore indentation
        //  Find the start of actual content
        while (comStart < length && IsParamBreak(str[comStart])){
            comStart++;
        }

        // Count the number of parameters
        unsigned long vecLen = 0;
        for (unsigned long i=comStart; i<length; i++){
            if ( IsParamBreak(str[i]) ){
                vecLen++;
            }
        }
        res.reserve(vecLen);


        // Scan will fail if there is no data
        if (length == 0){
            return res;
        }


        // Slice the the string into parts
        unsigned long toIndex = 0;
        unsigned long cursor = comStart;
        for (unsigned long i=cursor; i<length; i++){
            if (IsParamBreak(str[i]) == true){
                // Transfer the sub-string to the vector,
                //  Ensure that the data is it's own, and not a reference
                res[toIndex].reserve(i-cursor);

                // Error here
                res[toIndex].assign( str.substr(cursor, i-cursor) );

                cursor = i+1;
                toIndex++;
            }
        }

        return res;
    };


    StrCommands Fragment(std::string str){
        str.append("\n"); // Ensures that the loop will cover all segments
        unsigned long length = str.size();

        // Result
        StrCommands res;


        // Count lines
        //  Ignoring empty lines
        unsigned long vecLen = 1;
        for (unsigned long i=0; i<length; i++){
            if (IsLineBreak(str[i])){
                vecLen++;
            }
        }
        res.reserve(vecLen);


        // Ignore 'empty' strings as they may cause errors
        if (vecLen == 0){
            return res;
        }


        // Read lines
        unsigned long toIndex = 0;
        unsigned long cursor = 0;
        for (unsigned long i=0; i<length; i++){
            if (IsLineBreak(str[i])){

                // Error here
                res[toIndex].param = ParamArray(  Parameterize( std::string(str.substr(cursor, i-cursor)) )  );
                res[toIndex].line = i+1;

                // Ignore blank lines
                if (res[toIndex].param.size() == 0){
                    vecLen--;
                }else{
                    toIndex++;
                }
                cursor = i+1;
            }
        }


        // Shrink the result due to undersizing for blank lines
        res.reserve(vecLen);

        return res;
    };
}

内存访问冲突通常发生在第 66 和 108 行（当元素数据本地存储在向量中时）。它似乎发生在分配阶段，正如通过使用中间临时变量在解析后直接存储结果所推断的那样。该错误也可能在 vector::reserve() 期间发生，但发生频率较低。

注：在Windows上没有直接报错信息：

Exception thrown at 0x00A20462 in fiber.exe: 0xC0000005: Access violation reading location 0xBAADF009.

仅在使用 'C/C++ Extension for Visual Studio Code' 调试时可见，在正常终端执行中不可见。
然而在 Ubuntu 它输出：

Segmenation fault (core dump)

Answer 1

您在向量上调用 reserve，它分配内存来存储您的对象但不构造它们。当您随后尝试使用您尚未构建的对象的方法时，它很可能会崩溃。

有 2 种可能的解决方案，要么调用 resize 而不是 reserve，要么调用 push_back 在向量的末尾构造新对象。

C++：导致访问冲突的向量元素分配

C++: Vector element assignment causing Access violation

c++

vector

access-violation

c++14