OpenRTMFP/Cumulus Primer(5)CumulusServer启动流程分析(续)
ServerApplication 对其子类有如下要求:
Application 对其子类的要求是,如下这些成员函数必须被覆盖:
CumulusServer是继承ServerApplication的,ServerApplication是继承Application的。Application的run()函数会先调用initialize(),然后调用main(),最后调用uninitialize。最后这个反初始化过程,在CumulusServer就是直接调用父类的uninitialize函数。
void uninitialize() { ServerApplication::uninitialize();}CumulusServer的命令行选项有:log(l)、dump(d)、cirrus(c)、middle(m)、help(h)。
void defineOptions(OptionSet& options) { ServerApplication::defineOptions(options);设定日志级别(0 - 8,默认是 6,表示 info 级别)。
options.addOption( Option("log", "l", "Log level argument, must be beetween 0 and 8 : \ nothing, fatal, critic, error, warn, note, info, debug, trace. \ Default value is 6 (info), all logs until info level are displayed.") .required(false) .argument("level") .repeatable(false));其他一些选项:
options.addOption( Option("dump", "d", "Enables packet traces in logs. Optional arguments \ are 'middle' or 'all' respectively to displays just middle packet \ process or all packet process. If no argument is given, just outside \ packet process will be dumped.",false,"middle|all",false) .repeatable(false)); options.addOption( Option("cirrus", "c", "Cirrus address to activate a 'man-in-the-middle' \ developer mode in bypassing flash packets to the official cirrus \ server of your choice, it's a instable mode to help Cumulus developers, \ "p2p.rtmfp.net:10000" for example. By adding the 'dump' argument, \ you will able to display Cirrus/Flash packet exchange in your logs \ (see 'dump' argument).",false,"address",true) .repeatable(false)); options.addOption( Option("middle", "m","Enables a 'man-in-the-middle' developer mode \ between two peers. It's a instable mode to help Cumulus developers. \ By adding the 'dump' argument, you will able to display Flash/Flash \ packet exchange in your logs (see 'dump' argument).") .repeatable(false));显示帮助信息的选项:
options.addOption( Option("help", "h", "Displays help information about command-line usage.") .required(false) .repeatable(false));}OptionSet是Poco::Util::OptionSet,调用addOption可以向其中增加选项Option。其中required和repeatable表示:
当需要显示帮助信息时,调用如下函数:
void displayHelp() { HelpFormatter helpFormatter(options()); helpFormatter.setCommand(commandName()); helpFormatter.setUsage("OPTIONS"); helpFormatter.setHeader("CumulusServer, open source RTMFP server"); helpFormatter.format(cout);}参数是选项名和选项值。
void handleOption(const std::string& name, const std::string& value) { ServerApplication::handleOption(name, value);如果选项是帮助:
if (name == "help") _helpRequested = true;如果是cirrus,即该服务的 IP 和端口号,Poco::URI 中有协议名(Scheme)、IP 地址(Host)、端口号(Port)、查询串(Query)等等。
else if (name == "cirrus") { try { URI uri("rtmfp://"+value); _pCirrus = new SocketAddress(uri.getHost(),uri.getPort()); NOTE("Mode 'man in the middle' : the exchange will bypass to '%s'",value.c_str()); } catch(Exception& ex) { ERROR("Mode 'man in the middle' error : %s",ex.message().c_str()); }如果选项是dump日志:
} else if (name == "dump") { if(value == "all") Logs::SetDump(Logs::ALL); else if(value == "middle") Logs::SetDump(Logs::MIDDLE); else Logs::SetDump(Logs::EXTERNAL);如果选项是middle:
} else if (name == "middle") _middle = true;如果选项是log,表示设定日志级别:
else if (name == "log") Logs::SetLevel(atoi(value.c_str()));}先加一个作用域锁,然后再向日志流写数据。
void dumpHandler(const UInt8* data,UInt32 size) { ScopedLock<FastMutex> lock(_logMutex); cout.write((const char*)data, size); _logStream.write((const char*)data,size); manageLogFile();}调用 manageLogFile,主要做一些日志大小超出限制的处理。
void manageLogFile() {先判断是否超过日志文件的大小上线,LOG_SIZE是1000000字节(即约 1 MB)。
if(_pLogFile->getSize() > LOG_SIZE) { _logStream.close(); int num = 10;打开新日志文件:
File file(_logPath + "10");如果该文件已经存在,则先删除:
if (file.exists()) file.remove(); while (--num >= 0) { file = _logPath + NumberFormatter::format(num); if (file.exists()) file.renameTo(_logPath + NumberFormatter::format(num + 1)); } _logStream.open(_pLogFile->path(), ios::in | ios::ate); } }CumulusServer继承了ApplicationKiller,该类中有纯虚函数kill()需要被实现,于是有:
void kill() { terminate();}ApplicationKiller的定义在ApplicationKiller.h中,如下:
class ApplicationKiller {public: ApplicationKiller(){} virtual ~ApplicationKiller(){} virtual void kill()=0;};在initialize()函数中调用,上一篇已提到过。
void loadConfiguration(const string& path) { try { ServerApplication::loadConfiguration(path); } catch(...) { }}void logHandler(Thread::TID threadId, const std::string& threadName, Priority priority, const char *filePath, long line, const char *text) {作用域锁:
ScopedLock<FastMutex> lock(_logMutex); Path path(filePath); string file; if (path.getExtension() == "lua") file += path.directory(path.depth()-1) + "/";如果是命令行交互模式(即不是 daemon 模式):
if (_isInteractive) printf("%s %s[%ld] %s\n", g_logPriorities[priority - 1], (file + path.getBaseName()).c_str(), line, text);向日志流输出一句日志:
_logStream << DateTimeFormatter::format(LocalDateTime(),"%d/%m %H:%M:%S.%c ") << g_logPriorities[priority-1] << '\t' << threadName << '(' << threadId << ")\t" << (file + path.getFileName()) << '[' << line << "] " << text << std::endl; _logStream.flush();日志文件的善后处理(主要处理文件大小限制可能产生的问题):
manageLogFile();}-
转载请注明来自柳大的CSDN博客:Blog.CSDN.net/Poechant
-