question on order of headers and wexitstatus



in one of the unix programming faq's they have the following headers in the program to catch sigchld


code:

#include <sys/types.h>  /* include this before any other sys headers */
#include <sys/wait.h>   /* header for waitpid() and various macros */
#include <signal.h>     /* header for signal functions */
#include <stdio.h>      /* header for fprintf() */
#include <unistd.h>     /* header for fork() */

why does <sys/types.h> have to be placed before any other sys headers in this case?

and the second question. in the following code snippet, when i have status>>8 on a sep. line, i get an exit status of 256 when no one is logged on.



code:

if (status != 0) {
    fprintf(stderr, "no one logged on.\n");
    wexitstatus(status);
    status>>8;
    printf("%d\n", status);
    exit(exit_failure);
    }



however, when i put status>>8 in printf() like in the following

code:

if (status != 0) {
    fprintf(stderr, "no one logged on.\n");
    wexitstatus(status);
    printf("%d\n", status>>8);
    exit(exit_failure);
    }


i get an exit status of 1 when no one is logged on. i don't understand how the placement of status>>8 has an impact on the exit code.

q2 - posix says:

quote:
wexitstatus(stat_val)
if the value of wifexited(stat_val) is non-zero, this macro evaluates to the low-order 8 bits of the status argument that the child process passed to _exit() or exit(), or the value the child process returned from main().




the lower eight bits of something is never 256 -- 8 bits range from 0 - 255. first. did you check wifexited?

the >> question: i do not know, but i do know function arguments are sequence points. that means they have to be fully evaluated before they are passed as argument to a function.

are you using an old compiler, like the one that comes as part of hpux?

q1- sys/types.h is required to be included first by wait(), followed by sys/wait.h - if you read the man page for wait.

the real reason is so that test macros in the other include files do not "accidentally" turn off a declaration of some datatype, struct or typedef needed by somebody in another include file later on.

the simple answer is "sys/types.h is a fundamental header and contains many types required by other headers".

quote:
originally posted by jim mcnamara
q2 - posix says:


the lower eight bits of something is never 256 -- 8 bits range from 0 - 255. first. did you check wifexited?

the >> question: i do not know, but i do know function arguments are sequence points. that means they have to be fully evaluated before they are passed as argument to a function.

are you using an old compiler, like the one that comes as part of hpux?

q1- sys/types.h is required to be included first by wait(), followed by sys/wait.h - if you read the man page for wait.

the real reason is so that test macros in the other include files do not "accidentally" turn off a declaration of some datatype, struct or typedef needed by somebody in another include file later on.




i'm using the gnu c compiler. i'm suspecting it might have to do with that fact that the '>> isn't a sequence point.

1. you are not using the macros correctly...

it should be
code:
x=wexitstatus(status);

saying
code:
wexitstatus(status);by itself evaluates the exit status and discards the result.

2. similarly
code:
status>>8;effectively does nothing, it shifts a copy of the number right by 8 places and throws the result away.