内存分配失败的处理

看到effective c++中内存分配的处理，编写代码便于以后的理解：

1. 基本版：

#include <new.h> #include <iostream> using namespace std; int PrintError(size_t size) { cout<<"memory failed in allocation/t"<<size<<endl; return 0; //return 1; if executes here, then this function will be called //until the memory allocation success. } void main() { _set_new_handler(PrintError); int *pInt = new int[999999999999999]; } 

msdn上这么解释：

Call the C++ _set_new_handler function to specify an exception-handling function that is to gain control if the new operator fails to allocate memory. If new fails, the run-time system automatically calls the exception-handling function that was passed as an argument to _set_new_handler. _PNH, defined in NEW.H, is a pointer to a function that returns type int and takes an argument of type size_t. Use size_t to specify the amount of space to be allocated.

_set_new_handler is essentially a garbage-collection scheme. The run-time system retries allocation each time your function returns a nonzero value and fails if your function returns 0.

 

2. effective c++上提供的一条有效的建议是提供一个类作为基类，这样子类

等出现内存分配失败时，均可以用到，也可以写为模板类，代码如下：

 

NewClass.h代码： // NewClass.h: interface for the CNewClass class. // ////////////////////////////////////////////////////////////////////// #if !defined(AFX_NEWCLASS_H__90CDE415_B093_4E3C_B1B9_DD3DA750E7A9__INCLUDED_) #define AFX_NEWCLASS_H__90CDE415_B093_4E3C_B1B9_DD3DA750E7A9__INCLUDED_ #if _MSC_VER > 1000 #pragma once #endif // _MSC_VER > 1000 #include <new.h> class CNewClass { public: CNewClass(); virtual ~CNewClass(); public: static _PNH set_new_handler(_PNH p); static void * operator new(size_t size); private: static _PNH currenthandler; int m_int[9999999999]; }; #endif // !defined(AFX_NEWCLASS_H__90CDE415_B093_4E3C_B1B9_DD3DA750E7A9__INCLUDED_) NewClass.cpp文件代码： // NewClass.cpp: implementation of the CNewClass class. // ////////////////////////////////////////////////////////////////////// #include "NewClass.h" #include <new> ////////////////////////////////////////////////////////////////////// // Construction/Destruction ////////////////////////////////////////////////////////////////////// CNewClass::CNewClass() { } CNewClass::~CNewClass() { } _PNH CNewClass::set_new_handler(_PNH p) { _PNH oldhandler = currenthandler; currenthandler = p; return oldhandler; } void * CNewClass::operator new(size_t size) { _PNH globalhandler = _set_new_handler(currenthandler); void *memory; try { // 尝试分配内存 memory = ::operator new(size); } catch (std::bad_alloc&) { // 恢复旧的new_handler _set_new_handler(globalhandler); throw; // 抛出异常 } _set_new_handler(globalhandler); // 恢复旧的new_handler return memory; } _PNH CNewClass::currenthandler; 在main中使用代码： #include "NewClass.h" #include <stdlib.h> #include <iostream> using namespace std; int NonMemory(size_t size) { cout<<size<<"has been allocated!"<<endl; return 0; } void main() { CNewClass *pnew; CNewClass::set_new_handler(NonMemory); pnew = new CNewClass; pnew = new CNewClass; }  

为了让内存尽快耗完，我在class内部定义一个大block的内存块，这样就能在

main中定义到第二个变量时变出现了内存分配失败的情形，其实在NonMemory这样的函数中，

应该尽量做到垃圾内存回收，并且返回1，这样的话，会让分配失败后，继续调用分配内存，这样才是

我们编程人员想要的结果。以上纯为effective c++中“预先准备内存不足”这一章节的个人理解描述。