当std :: ofstream超出范围时,我很少遇到seg错误问题。我在Logger.h中有以下结构:
struct LogRequest
{
std::string line;
std::string prefix;
};
struct FileInfo
{
std::string name;
std::ofstream ofs;
int lines;
std::string ToString()
{
return (name + " exists");
};
在Logger.cpp中,我有以下功能:
void Logger::Log(const std::string line, std::string prefix)
{
pthread_mutex_lock(&mutex);
if (file_map.find(prefix) == file_map.end())
{
OpenLogFile(prefix);
}
LogRequest* request = new LogRequest;
request->line = line;
request->prefix = prefix;
message_queue.push_back(request);
pthread_mutex_unlock(&mutex);
}
void Logger::OpenLogFile(const std::string& prefix)
{
//get timestamp to use
char timestamp[15];
time_t now;
time(&now);
struct tm* current = localtime(&now);
sprintf(timestamp, "%02u%02u%04u_%02u%02u%02u", (current->tm_mon+1),
current->tm_mday,(1900 + current->tm_year), current->tm_hour,
current->tm_min, current->tm_sec);
FileInfo* info = new FileInfo;
info->name = "logs/" + prefix + ".log_" + timestamp;
info->ofs.open(info->name.c_str(), std::ios::out);
info->lines = 0;
file_map[prefix] = info;
}
void Logger::CloseLogFile(const std::string& prefix)
{
delete file_map[prefix];
}
在Logger.cpp的一个帖子中,我有......
void Logger::WriteToFile()
{
std::map<std::string, FileInfo* >::iterator it;
while(run)
{
char timestamp[16];
time_t now;
time(&now);
struct tm* current = localtime(&now);
sprintf(timestamp, "%02u%02u%04u|%02u%02u%02u|", (current->tm_mon+1),
current->tm_mday,(1900 + current->tm_year), current->tm_hour,
current->tm_min, current->tm_sec);
pthread_mutex_lock(&mutex);
for(it=file_map.begin(); it != file_map.end(); ++it)
{
if(it->second->lines > MAX_LINES)
{
CloseLogFile(it->first);
OpenLogFile(it->first);
}
else
{
int written = 0;
while(!message_queue.empty() && written < MESSAGES_PER_WRITE)
{
LogRequest* request = message_queue.front();
message_queue.pop_front();
std::string line(timestamp, 16);
line.append(request->line);
FileInfo* info = file_map[request->prefix];
info->ofs << line << std::endl;
info->lines++;
written++;
delete request;
}
}
}
pthread_mutex_unlock(&mutex);
usleep(1000);
}
}
我遇到的问题是,当我尝试CloseLogFile
时,有一个seg错误会在FileInfo中的std :: ofstream ofs的析构函数中抛出。它工作很多次。甚至在SEG Fault发生之前(通过各种cout语句)我已经确认std :: ofstream是好的,没有失败,没有达到eof,并且还不错。我还确认file_map [prefix]存在,并且通过输出FileInfo的ToString()也存在。这些都是在CloseLogFile中删除之前检查的。
我的问题涉及SEG FAULT的原因。
当行在WriteToFile()中输出到ofstream时,如果删除LogRequest对象,则可以。如何,info->ofs << line << std::endl;
行中究竟发生了什么?
答案 0 :(得分:1)
看起来像你的sprintf(timestamp ...)调用被一个字符(null)覆盖,这将导致未定义的行为。可能是也可能不是你的主要问题,因为那是在一个线程堆栈上,你的ons在堆上......