通过TCP传输文件末尾多余空行

2 投票
1 回答
799 浏览
提问于 2025-04-15 23:09

我有两个程序,一个是recvfile.py,另一个是sendfile.cpp。它们基本上能正常工作,但我发现新文件的末尾多了很多空行。我不知道这些多余的空行是怎么来的。我知道问题出在发送端,因为当我用Python的sendall()函数发送文件时,就不会出现这种情况。

以下是这些文件:

jmm_sockets.c

#include <winsock.h>
#include <stdio.h>
#include <stdlib.h>

int getServerSocket(int port)
{
  WSADATA wsaData;
  if(WSAStartup(MAKEWORD(2,0), &wsaData) != 0){
    fprintf(stderr, "WSAStartup() failed\n");
    exit(1);
  }

  // create socket for incoming connections
  int servSock = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
  if(servSock == INVALID_SOCKET){
    fprintf(stderr, "Oops: socket() failed %d\n", WSAGetLastError());
    exit(1);
  }

  // construct local address structure
  struct sockaddr_in servAddr;
  memset(&servAddr, 0, sizeof(servAddr));
  servAddr.sin_family = AF_INET;
  servAddr.sin_addr.s_addr = INADDR_ANY;
  servAddr.sin_port = htons(port);

  // bind to the local address
  int servAddrLen = sizeof(servAddr);
  if(bind(servSock, (SOCKADDR*)&servAddr, servAddrLen) == SOCKET_ERROR){
    fprintf(stderr, "Oops: bind() failed %d\n", WSAGetLastError());
    exit(1);
  }

  return servSock;
}

int getClientSocket(char* host, int port)
{
  WSADATA wsaData;
  if(WSAStartup(MAKEWORD(2,0), &wsaData) != 0){
    fprintf(stderr, "Oops: WSAStartup() failed");
    exit(1);
  }

  // create tcp socket
  int sock = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
  if(socket<0){
    fprintf(stderr, "Oops: socket() failed %d\n", WSAGetLastError());
    exit(1);
  }

  // set up serverAddr structure
  struct sockaddr_in servAddr;
  memset(&servAddr, 0, sizeof(servAddr));
  servAddr.sin_family = AF_INET;
  servAddr.sin_addr.s_addr = inet_addr(host);
  servAddr.sin_port = htons(port);

  // connecet to server address
  if(connect(sock, (SOCKADDR*)&servAddr, sizeof(servAddr)) < 0){
    fprintf(stderr, "Oops: connect() failed. %d\n", WSAGetLastError());
    exit(1);
  }

  return sock;
}

sendfile.cpp:

#include "jmm_sockets.h"
#include <windows.h>
#include <iostream>
#include <fstream>
#include <cstdlib>
#include <sys/stat.h>
using namespace std;

int main(int argc, char** argv)
{
  int port;
  string host;
  string filename;

  if(argc==2){
    cout << "Host: ";
    cin >> host;

    cout << "Port: ";
    cin >> port;

    filename = argv[1];
  }else if (argc == 4){
    host = argv[1];
    port = atoi(argv[2]);
    filename = argv[3];
  }else{
    cerr << "Usage: " << argv[0] << " [<host> <port>] <filename>" << endl;
    exit(1);
  }

  // open file for reading
  ifstream fin;
  fin.open(filename.c_str());
  if(fin.fail()){
    cerr << "Error: opening " << filename << " failed. " << endl;
    exit(1);
  }

  // get file size
  fin.seekg(0, ios::end);
  int size = fin.tellg();
  fin.seekg(0, ios::beg);

  // open socket for sending
  int sock = getClientSocket((char*)host.c_str(), port);

  // send file size
  char buffer[16];
  itoa(size, buffer, 10);
  int i;
  for(i=0; i<strlen(buffer); i++){
    if(send(sock, &buffer[i], 1, 0)!=1){
      cerr << "Error: send() failed " << WSAGetLastError() << endl;
      exit(1);
    }
  }
  char c = '\n';
  if(send(sock, &c, 1, 0)!=1){
    fprintf(stderr, "Error: send() failed %d\n", WSAGetLastError());
    exit(1);
  }

  // recv y or n
  int recvMsgSize = recv(sock, &c, 1, 0);
  if(recvMsgSize!=1){
    fprintf(stderr, "Error: recv() failed %d\n", WSAGetLastError());
    exit(1);
  }

  if(c=='y'){
    // send entire file
    int readSoFar = 0;
    while(readSoFar < size){
      fin.get(c);
      if(send(sock, &c, 1, 0)!=1){
    cerr << "Error: send() failed " << WSAGetLastError() << endl;
    exit(1);
      }
      readSoFar++;
    }

  }else if (c=='n'){
    // leave
    cout << "Remote host declined file." << endl;
  }

  fin.close();
  closesocket(sock);
  WSACleanup();

  // 
  return 0;
}

最后是recvfile.py:

import sys
from jmm_sockets import *
import yesno

if len(sys.argv) != 2:
    print "Usage: ", argv[0], "<port>"

s = getServerSocket(None, int(sys.argv[1]))
conn, addr = s.accept()

buffer = None
filelen = str()

# receive filesize
while 1:
    buffer = conn.recv(1)
    if buffer == '\n':
        # give it a rest
        break
    else:
        filelen = filelen + buffer

# prompt user to accept file
filelen = int(filelen)
print "file size = ", filelen,
userChoice = yesno.yesno("Accept?")
conn.send(userChoice)

# conditionally accecpt file
if bool(userChoice):
    filename = raw_input("What do you want to call the file? ")
    f = open(filename, 'w')

    buffer = None
    data = str()
    recvdBytes = 0
    while recvdBytes < filelen:
        buffer = conn.recv(1)
        recvdBytes = recvdBytes + 1
        data = data + buffer

print "File: ",
f.write(data)
print "written"
conn.close()

1 个回答

2

你遇到多余换行符的原因是因为你在通过套接字发送了多余的换行符,这主要是因为你试图发送的数据量超过了应该发送的量。

如果你检查一下输入文件 finfail() 状态,你会发现它在最后几次调用 fin.get(c) 时失败了,因此 c 的值没有改变——它仍然是换行符,这是输入文件中的最后一个字符。

这种情况发生的原因是因为 CRLF 转换:你使用的文件大小(size 变量)是磁盘上原始的文件大小,计算了所有的 CR(回车符)。但是,当你以文本模式打开文件并逐字节读取时,标准库会默默地将所有的 CRLF 转换成 LF(换行符),所以你并没有通过套接字发送 CR。因此,你在这个过程中得到的额外换行符数量等于原始文件中换行符的数量。

解决这个问题的方法是以二进制模式打开文件,这样就可以禁用 CRLF 转换:

fin.open(filename.c_str(), ios::in | ios::binary);

此外,你不应该逐字节发送文件——这样速度非常慢。如果运气不好,你可能会为每个字节发送一个完整的数据包。如果运气好,你的操作系统的网络栈会将这些多个发送合并成更大的数据包(但不要依赖于这一点),即使这样,你仍然会进行大量的系统调用。

考虑重构你的代码,减少对 send()recv() 的调用次数,尽量每次传递更多的字节,例如:

// Just use one call to send instead of looping over bytes and sending one
// byte at a time.  Simpler and faster!
send(sock, buffer, strlen(buffer), 0);

撰写回答