printf输出条件：
(1) 调用fflush；
(2) 缓冲区满了；
(3) 遇到 这些字符
(4) 遇到scanf这些要取缓冲区的；
(5) 线程或者进程退出；
fork之后会拷贝父进程的缓冲区；
代码：
1 #include <unistd.h> 2 #include <stdio.h> 3 4 5 int glob = 6; 6 char buf[ ] = "a write to stdout "; 7 8 int main() 9 { 10 int var; 11 pid_t pid; 12 13 var = 88; 14 15 if (write(STDOUT_FILENO, buf, sizeof(buf) - 1) != sizeof(buf) - 1){ 16 perror("write error "); 17 return -1; 18 } 19 20 printf("before fork "); 21 22 if ((pid = fork()) < 0){ 23 perror("fork error "); 24 return -1; 25 } else if (pid == 0){ 26 glob++; 27 var++; 28 } else { 29 sleep (2); 30 } 31 32 printf("pid = %d, glob = %d, var = %d ", getpid(), glob, var); 33 34 return 0; 35 }
结果：
[root@AY1404041716143649d3Z forkPrintf]# gcc forkPrintf.c -o forkPrintf [root@AY1404041716143649d3Z forkPrintf]# ./forkPrintf a write to stdout before fork pid = 17098, glob = 7, var = 89 pid = 17097, glob = 6, var = 88 [root@AY1404041716143649d3Z forkPrintf]# ./forkPrintf > tmp.out [root@AY1404041716143649d3Z forkPrintf]# cat tmp.out a write to stdout before fork pid = 17106, glob = 7, var = 89 before fork pid = 17105, glob = 6, var = 88
(1) 第一次输入到控制台，write无缓冲直接输出，printf("before fork ")存在 行缓冲标识直接输出，而子进程拷贝父进程代码段，随后也会输出最后的printf;
(2) 第二次输入到文件，write无缓冲直接输出，printf("before fork ")成了全缓冲的并没有输出，子进程拷贝了父进程的缓冲区和代码段，所以当父子进程退出时，刷新缓冲区，均有两条printf输出；