When I run your code verbatim (well, OK; I added void
to each empty argument list and made the child functions static
), then I got sample output:
main pid is:46761
my pid is:46762
I am child1..
my pid is:46763
my pid is:46764
I am child1..
I am child2.., my parent is:46762
my pid is:46765
I am child2.., my parent is:1
my pid is:46766
I am child1..
my pid is:46768
my pid is:46767
I am child2.., my parent is:46766
I am child1..
my pid is:46769
I am child2.., my parent is:1
For amusement value, I ran the output to a pipe and got:
main pid is:46770
main pid is:46770
my pid is:46773
I am child1..
main pid is:46770
my pid is:46772
I am child1..
main pid is:46770
my pid is:46775
I am child2.., my parent is:1
main pid is:46770
my pid is:46774
I am child2.., my parent is:46772
main pid is:46770
my pid is:46772
I am child1..
my pid is:46776
I am child1..
main pid is:46770
my pid is:46772
I am child1..
my pid is:46778
I am child2.., my parent is:1
main pid is:46770
my pid is:46774
I am child2.., my parent is:46772
my pid is:46777
I am child1..
main pid is:46770
my pid is:46774
I am child2.., my parent is:46772
my pid is:46779
I am child2.., my parent is:46777
See printf()
anomaly after fork()
for an explanation of that behaviour.
Then I instrumented your code slightly differently, but the core logic is unchanged:
#include <stdio.h>
#include <sys/wait.h>
#include <unistd.h>
static void child1(void)
{
printf("Child1: PID = %d, PPID = %d\n", (int)getpid(), (int)getppid());
}
static void child2(void)
{
printf("Child2: PID = %d, PPID = %d\n", (int)getpid(), (int)getppid());
}
int main(void)
{
pid_t childpid[4];
int i;
printf("main pid is:%d\n", getpid());
for (i = 0; i < 2; i++)
{
fflush(0);
childpid[i] = fork();
if (childpid[i] < 0)
printf("fork fail in PID %d\n", getpid());
else if (childpid[i] == 0)
{
childpid[i + 2] = fork();
if (childpid[i + 2] < 0)
printf("fork fail in PID %d\n", getpid());
else if (childpid[i + 2] == 0)
child2();
else
{
printf("child pid %d forked child %d\n", (int)getpid(), (int)childpid[i+2]);
child1();
}
}
else
printf("main pid %d forked child %d\n", (int)getpid(), (int)childpid[i]);
}
int status;
int corpse;
while ((corpse = wait(&status)) != -1)
printf("PID %d: child %d died with status %.4X\n", (int)getpid(), corpse, status);
printf("PID %d: finished\n", (int)getpid());
return getpid() % 256;
}
I did add the code to wait for children to die, and to report when a process finishes, and added non-zero exit statuses for most of the processes. The code also flushes standard output before forking to keep things clean (even if the output is going to a pipe or file).
On my Mac OS X 10.11.1 El Capitan machine, I got, for example:
main pid is:46730
main pid 46730 forked child 46731
main pid 46730 forked child 46732
child pid 46731 forked child 46733
Child1: PID = 46731, PPID = 46730
child pid 46732 forked child 46734
main pid 46731 forked child 46735
Child1: PID = 46732, PPID = 46730
Child2: PID = 46733, PPID = 46731
Child2: PID = 46734, PPID = 46732
PID 46734: finished
main pid 46733 forked child 46736
PID 46732: child 46734 died with status 8E00
PID 46732: finished
child pid 46735 forked child 46737
Child1: PID = 46735, PPID = 46731
PID 46730: child 46732 died with status 8C00
Child2: PID = 46737, PPID = 46735
child pid 46736 forked child 46738
Child1: PID = 46736, PPID = 46733
PID 46737: finished
Child2: PID = 46738, PPID = 46736
PID 46735: child 46737 died with status 9100
PID 46735: finished
PID 46738: finished
PID 46736: child 46738 died with status 9200
PID 46731: child 46735 died with status 8F00
PID 46736: finished
PID 46733: child 46736 died with status 9000
PID 46733: finished
PID 46731: child 46733 died with status 8D00
PID 46731: finished
PID 46730: child 46731 died with status 8B00
PID 46730: finished
When you study the sequencing in that output, there are some interesting interleavings of execution.
None of that explains why you got 1118; in many ways, that is impossible to explain because no-one can sequence things exactly as on your machine. However, what I've shown is plausible behaviour, and it does not reproduce the 'all processes have a single PID as the PPID'. You should be able to make incremental monitoring improvements to your code and demonstrate which change alters the behaviour on your machine.