在尝试构建一个可以使用boost async_write
传输数据块/数组的程序时,我的线程遇到了麻烦:
这就是我想要执行的内容:
write(unsigned char *pMsg, unsigned short nMsgLen){
io_service_.post(boost::bind(&m_client::write_buf, this, pMsg, nMsgLen));
}
write_buf(unsigned char *pMsg, unsigned short nMsgLen){
boost::asio::async_write(target,
boost::asio::buffer(pMsg, nMsgLen),
boost::bind(&m_client::write_buf_compl,
this,
boost::asio::placeholders::error));
}
它确实编译,但是pMsg
在调用write_buf
时没有正确的内容,我认为这是因为它不在同一个线程内部调用。
那么,如何调整此构造以将我的数组作为参数传输?!
答案 0 :(得分:0)
您似乎非常了解问题所在。快速解决方法是通过复制其内容来传递缓冲区:
void write(unsigned char *pMsg, unsigned short nMsgLen) {
// TODO: Take care of exception safety...
char *pMsgCopy = (char *)malloc(nMsgLen);
memcpy(pMsgCopy, pMsg, nMsgLen);
io_service_.post(boost::bind(&m_client::write_buf, this,
pMsgCopy, nMsgLen));
}
void write_buf(unsigned char *pMsg, unsigned short nMsgLen)
{
// Achtung! Don't forget to free memory in complection callback.
boost::asio::async_write(target,
boost::asio::buffer(pMsg, nMsgLen),
boost::bind(&m_client::write_buf_compl,
this,
boost::asio::placeholders::error));
}
void write_buf_compl(...)
{
free(pMsg);
}
如果您对效率感到痴迷,那么您可以通过让write()
的调用者首先提供动态分配的缓冲区来避免复制。但是,如果你真的对性能感到疯狂,那么我建议不要使用Boost.ASIO。但这是另一个故事。