C++memset内存陷阱

今日接到任务，公司一套UNIX下的C++程序每隔一段时间会出现僵死，处理变慢，重复处理数据的情况。这种情况延续了2年之久，每次出问题都是由运维人员进行手动重启处理，由于涉及到电信用户的计费，该情况无法复现，现在需要我来定位该困境。

该程序在用户态24小时运行，整体架构使用C++的面向对象特性，由一个基类继承，使用多态实现多个子类各自独立的业务处理，采用多进程单线程的模式。近日刚好碰到一次程序处理变慢并重复处理数据的情况，对日志进行了详细的截取分析，将ACE的ERROR日志，TRC日志结合起来，定位了处理函数退出的地方，将重启后的处理成功日志进行对比，将问题定位到类DCCustPriceSyn的GetAllPricePlan()成员函数中。由于这个程序经历了不同程序员的手，代码变得比较杂乱，并且没有catch到异常的地方，只在顶层catch了一个unknow 异常。

下面贴一段模块测试的代码

#include <stdio.h>
#include <stdlib.h>
#include <iostream>
#include <list>
#include <map>
#include <vector>
#include <math.h>
using namespace std;

#define OTL_ORA10G
#include "otlv4.h"
#define OTL_BIGINT  __int64
#define OTL_STR_TO_BIGINT(str,n) \
{ \
  n=_atoi64(str); \
}

#define OTL_BIGINT_TO_STR(n,str) \
{ \
  _i64toa(n,str,10); \
}

const char* my_pConnectOcs = "query_ods/avpx09@hbocs";


struct STBindInfo
{
    char ObjType[4];
    long ObjID;
};


struct STServEffInfo
{
	long lnServID;
	char EffDate[24];
	long OfrDetailID;
	int nFlag; //flag = 0  说明此种产品实例用户未订购过
	STServEffInfo()
	{
		lnServID = 0;
		memset(EffDate,0,sizeof(lnServID));
		OfrDetailID = -1;
		nFlag = 0;
	}
};


struct STCustPlan
{
    STCustPlan()
    {
        ExpDate[0] = 0;
        ModifyTime[0] = 0;
        SourceType = 2;
        lnPrdId = -1;
        lnOfrType =1;
        OfrDetailID =0;
    }

    long CustPlanID;
    long CustPlanSeqID;
    long CustID;
    long PringPlanID;
    long OfrID;
    long SourceType;
    char State[4];
    long Priority; 
    char ExpDate[24];    
    char EffDate[24];
    char ModifyTime[24];
    long RegionID;
    long BelongObjID;
    long lnPricingLev;
    char BelongObjType[5];
    long lnPrdId;
    long lnOfrType;
	long lnOperSerialNbr;
    long OfrDetailID;
    list<STBindInfo> lBindInfo;
    vector<STServEffInfo> vecServEffInfo;
};








class DCOra
{
public:
	DCOra();
	~DCOra();
public:
	otl_connect db_ocs;
	int m_nPriority;
public:
	int GetAllPricePlan(long BelongObjID, list<STCustPlan>& lAllPricePlan);
	
	int getOfrPriority(long lnPricingPlanId, char *szBelongObjType, long& lnOfrId, long& lnPriority);
	void CheckDb()
	{



		char sqlCheck[256] = "\0";
		int connCheck = 0;
		otl_stream osCheck;
		strcpy(sqlCheck," select 1 from dual ");
		try
		{
			osCheck.open(1,sqlCheck,db_ocs);
			osCheck.close();
			connCheck = 1;

		}
		catch(otl_exception &p)
		{
			printf("::CheckDbConn", __FILE__, __LINE__,  " p.stm_text:%s--p.msg:%s--p.code:%d \n",p.stm_text,p.msg,p.code);
		}

		//重连数据库
		try
		{	
			switch(connCheck)
			{
				case 1:
				{
					db_ocs.logoff();
					db_ocs.rlogon(my_pConnectOcs);
					db_ocs.auto_commit_off();
					break;
				}
				default:
					break;
			}
		}
		catch(otl_exception& p)
		{
			printf("::CheckDbConn", __FILE__, __LINE__,  " p.stm_text:%s--p.msg:%s--p.code:%d \n",p.stm_text,p.msg,p.code);
		}
	}
};

DCOra::DCOra()
{
	try
	{
		otl_connect::otl_initialize(1); // initialize OCI environment
		db_ocs.rlogon(my_pConnectOcs);

		db_ocs.auto_commit_off();
	}
	catch(otl_exception& p)
	{
		// intercept OTL exceptions
		printf( "::DCRatable", __FILE__, __LINE__,	"Connect Error: (%s) (%s) (%s)\n",p.msg, p.stm_text, p.var_info);
	
	}

}
DCOra::~DCOra()
{
	db_ocs.commit();

	db_ocs.logoff();
}



int DCOra::GetAllPricePlan(long BelongObjID, list<STCustPlan>& lAllPricePlan)
{
	cout<<"GetAllPricePlan:Begin to get all price plan according :"<<BelongObjID<<endl;
		

	
		char my_str[1024];
		otl_stream os;
		lAllPricePlan.clear();
		STCustPlan sTempPlan;
		STServEffInfo sServEffInfo;
		int nRet;
		long lnTmpPriority = 0;
		
		strcpy(my_str," select CUST_PRICE_PLAN_ID,CUST_PRICE_PLAN_SEQ_ID,CUST_ID,PRICING_PLAN_ID,BELONG_OBJECT_TYPE,to_char(EFF_DATE),to_char(EXP_DATE), "
			" STATE,PARTITION_ID_REGION,to_char(MODIFY_TIME),SEQ_ID from ls65_sid.cust_price_plan_t@to_sid "
			" where BELONG_OBJECT_ID = :A<long> order by CUST_PRICE_PLAN_ID,CUST_PRICE_PLAN_SEQ_ID DESC ");
		os.open(1,my_str,db_ocs);
		os << BelongObjID;
		cout<<my_str<<endl;
		for( ;os.eof() == 0; )
		{
			memset(&sTempPlan, 0, sizeof(sTempPlan));
			os >> sTempPlan.CustPlanID >> sTempPlan.CustPlanSeqID >> sTempPlan.CustID >> sTempPlan.PringPlanID >> sTempPlan.BelongObjType >> sTempPlan.EffDate >> sTempPlan.ExpDate >> sTempPlan.State >> sTempPlan.RegionID >> sTempPlan.ModifyTime >> lnTmpPriority;

			sTempPlan.BelongObjID = BelongObjID;
			cout<<"$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$"<<endl;
			
			printf("eff_date is:%s!\n", sTempPlan.EffDate);
			
			printf("exp_date is:%s!\n", sTempPlan.ExpDate);
			cout<<"$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$"<<endl;
			
			if (strcmp(sTempPlan.BelongObjType, "80C") == 0)
			{
				cout<<"into 80C"<<endl;
				return 0;
			}

			if (strcmp(sTempPlan.BelongObjType, "80A") == 0)
			{
				strcmp(sServEffInfo.EffDate, sTempPlan.EffDate);
			}
			sTempPlan.vecServEffInfo.push_back(sServEffInfo);

			list<STCustPlan>::iterator iter = lAllPricePlan.begin();

			nRet = getOfrPriority(sTempPlan.PringPlanID, sTempPlan.BelongObjType, sTempPlan.OfrID, sTempPlan.Priority);
			if (nRet < 0)
			{
				printf("getOfrPriority PringPlanID =%d is failed", sTempPlan.PringPlanID);
				return nRet;
			}

			if (m_nPriority == 1)//seq_id覆盖
			{
				sTempPlan.Priority = lnTmpPriority;
			}


			for (; iter != lAllPricePlan.end(); ++iter)
			{
				if (sTempPlan.CustPlanID == iter->CustPlanID)
				{
					break;
				}
			}
			if (iter == lAllPricePlan.end())
				lAllPricePlan.push_back(sTempPlan);
		}


	printf("GetAllPricePlan:get all price plan according:%ld.it has %d price plan!", BelongObjID, lAllPricePlan.size());
	os.close();
	return 0;
}



int DCOra::getOfrPriority(long lnPricingPlanId, char *szBelongObjType, long& lnOfrId, long& lnPriority)
{
	printf("DCCustPriceSyn::getOfrPriority QuerySyn1508 PricingPlanId = %ld ", lnPricingPlanId);


		
		char my_str[1024];
		otl_stream os;
		strcpy(my_str,"select a.ocs_ofr_id ,c.CALC_PRIORITY from mid_pricing_plan_t a "
			" ,tb_prd_ofr_detail b,tb_prd_ofr c where a.sid_pricing_plan_id=:A<long> and a.ocs_ofr_id = b.ofr_id and b.ofr_id = c.ofr_id ");
		os.open(1,my_str,db_ocs);
		os << lnPricingPlanId;
		for( ;os.eof() == 0; )
		{
			os >> lnOfrId >> lnPriority;
		}
	


		if (strcmp(szBelongObjType, "80A") == 0)
		{
			lnPriority += 600000;
		}
		else if (strcmp(szBelongObjType, "80I") == 0)
		{
			lnPriority += 400000;
		}
		else if (strcmp(szBelongObjType, "80J") == 0)
		{
			lnPriority += 500000;
		}
		else if (strcmp(szBelongObjType, "80R") == 0)
		{
			lnPriority += 300000;
		}
		else
		{
			lnPriority += 600000;
		}

		printf("OfrID = %ld,Priority=%d\n", lnOfrId, lnPriority);

	os.close();
	return 0;
}





int main()
{
	DCOra *p = new DCOra();
	list<STCustPlan> lAllPricePlan;
	p->CheckDb();
	//while(TRUE)
	//{
		p->GetAllPricePlan(19042175676,lAllPricePlan);
	//}
	delete p;
	p = NULL;
	return 0;
}

业务逻辑处理是从程序中粘贴而来，不难看出，业务处理中有几处不当。结构体sServEffInfo并没有赋值，却被传入外部进行其他操作

			if (strcmp(sTempPlan.BelongObjType, "80A") == 0)
			{
				strcmp(sServEffInfo.EffDate, sTempPlan.EffDate);
			}

这段中，进入if后却进行了字符串比较操作，这显然是前任作者的笔误，按照逻辑此处应该是strcpy赋值操作。通过分析日志，发现执行到

printf("OfrID = %ld,Priority=%d\n", lnOfrId, lnPriority);

之后，程序便抛出了无法识别的异常，而且没有打印GetAllPricePlan:get all price plan according...，这说明了，程序没有退出while循环就抛异常了，这个地方大概在：

			if (nRet < 0)
			{
				printf("getOfrPriority PringPlanID =%d is failed", sTempPlan.PringPlanID);
				return nRet;
			}

			if (m_nPriority == 1)//seq_id覆盖
			{
				sTempPlan.Priority = lnTmpPriority;
			}


			for (; iter != lAllPricePlan.end(); ++iter)
			{
				if (sTempPlan.CustPlanID == iter->CustPlanID)
				{
					break;
				}
			}
			if (iter == lAllPricePlan.end())
				lAllPricePlan.push_back(sTempPlan);

或者：

			memset(&sTempPlan, 0, sizeof(sTempPlan));
			os >> sTempPlan.CustPlanID >> sTempPlan.CustPlanSeqID >> sTempPlan.CustID >> sTempPlan.PringPlanID >> sTempPlan.BelongObjType >> sTempPlan.EffDate >> sTempPlan.ExpDate >> sTempPlan.State >> sTempPlan.RegionID >> sTempPlan.ModifyTime >> lnTmpPriority;

			sTempPlan.BelongObjID = BelongObjID;

这个时候重点来了，我们注意到

memset(&sTempPlan, 0, sizeof(sTempPlan));

这个