union wait problem

zaichik · 12-21-2004, 10:07 PM

Hi all,

I'm trying to code a skeleton daemon process on Red Hat 9, using 14-year-old code snippets based on UNIX 4.3BSD and System V.

I am getting a couple of compile errors in the following function:

Code:

void sig_child() {
              
        int             pid;
        union wait      status;

        while( (pid = wait3( &status, WNOHANG, (struct rusage *) 0 )) > 0 );
}

The errors:

skel2.c:14: storage size of `status' isn't known
skel2.c:16: warning: implicit declaration of function `wait3'

I assume I am omitting some header files, but am unsure which. Here are my includes:

#include <stdlib.h>
#include <sys/types.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <sys/wait.h>
#include <unistd.h>
#include <stropts.h>

Thanks in advance!

itsme86 · 12-21-2004, 10:38 PM

Code:

NAME
       wait3, wait4 - wait for process termination, BSD style

SYNOPSIS
       #define _USE_BSD
       #include <sys/types.h>
       #include <sys/resource.h>
       #include <sys/wait.h>

       pid_t wait3(int *status, int options,
             struct rusage *rusage)

It looks like before you include your header files you need to #define _USE_BSD

And I'm not sure what union wait is. wait3() expects an int * for status.

zaichik · 12-22-2004, 09:36 AM

Hi, thanks for your post.

My man page on wait3 is a bit different:

SYNOPSIS
#include <sys/types.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <sys/wait.h>

pid_t wait3(int *status, int options,
struct rusage *rusage);

It doesn't specify #define __USE_BSD

I tried that anyway and it still didn't work. Although, in <sys/wait.h> there is a section that defines a union wait #ifdefined __USE_BSD. Curious.

At any rate, since wait3 is clearly expecting an int* for the first argument, I should be fine if I just declare

Code:

int status;

instead of

Code:

union wait status;

since (at this juncture) I am not concerned with portability. Does that makes sense?

P.S. What's up with all the leading underscore characters? Do they have some special significance by convention, or what?

itsme86 · 12-22-2004, 09:59 AM

Yes, I wouold do int status and then pass &status to wait3().

The C standard specifies that any variable/macro name that is all capital letters and begins with any number of leading underscores is reserved for implementation namespace. So gcc knows that it's safe (according to the standard) if it names its macros with leading underscores.

Who knows? Maybe someone wants to use a macro called USE_BSD in their own program

There's all sorts of namespace restrictions, but I don't know them all of the top of my head. At any rate, the leading underscores are just to avoid namespace contamination