I'm taking a class on Internet Programming, and the lab I'm currently
working on requires us to use threads and semaphores.  I'm looking at
a slide from that class titled "Unix Semaphore Code" that says:

[] Creation
o union semun arguments;
o key_t key = 1;
o int flags = 0777 | IPC_CREAT;
o int semid = semget(key, 1, flags);
o argument.val = initialvalue;
o semctl(semid, 0, SETVAL, argument);
[] Destruction
o int ignored_int;
o union semun ignored;
o semctl(semid, ignored_int, IPC_RMID, ignored);

I wanted to create an array of two semaphores, so I wrote the follow-
ing <main> function:

#include <pthread.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/ipc.h>
#include <sys/sem.h>

..

int main (  int   argCount
, char** arguments)
{
pthread_t Write[ 2];
union semun argument;               <=================
key_t key = 1;
int flags = 0700 | IPC_CREAT;
int before[ 2];
int pairs[ 2][ 2];
int result;
int parity;
int end;
printf( "Before creation of semaphores.\n");
for (parity = EVEN; parity <= ODD; parity++)
{ before[ parity] = semget( key, 1, flags);
argument.val = parity;
semctl( before[ parity], 0, SETVAL, argument);
}
..

But when I try to compile it I get the error message
"CountSharer.c:44: error: storage size of 'argument' isn't known".
Line 44 is the line my arrow is pointing to up above.  The slide up
above is the only reference to <union semun> that I know of.  Does
anyone know what I need to do to my <argument> variable to keep this
error from occurring?

---Kevin Simonson

"You'll never get to heaven, or even to LA,
if you don't believe there's a way."
from _Why Not_

[ Post a follow-up to this message ]

kvnsmnsn@hotmail.com writes:

> I'm taking a class on Internet Programming, and the lab I'm currently
> working on requires us to use threads and semaphores.  I'm looking at
> a slide from that class titled "Unix Semaphore Code" that says:
>
>      [] Creation
>         o union semun arguments;
>         o key_t key = 1;
>         o int flags = 0777 | IPC_CREAT;
>         o int semid = semget(key, 1, flags);
>         o argument.val = initialvalue;
>         o semctl(semid, 0, SETVAL, argument);
>      [] Destruction
>         o int ignored_int;
>         o union semun ignored;
>         o semctl(semid, ignored_int, IPC_RMID, ignored);
>
> I wanted to create an array of two semaphores, so I wrote the follow-
> ing <main> function:
>
>      #include <pthread.h>
>      #include <stdio.h>
>      #include <sys/types.h>
>      #include <sys/ipc.h>
>      #include <sys/sem.h>
>
>      ...
>
>      int main (  int   argCount
>               , char** arguments)
>      {
>        pthread_t Write[ 2];
>        union semun argument;               <=================
>        key_t key = 1;
>        int flags = 0700 | IPC_CREAT;
>        int before[ 2];
>        int pairs[ 2][ 2];
>        int result;
>        int parity;
>        int end;
>        printf( "Before creation of semaphores.\n");
>        for (parity = EVEN; parity <= ODD; parity++)
>        { before[ parity] = semget( key, 1, flags);
>          argument.val = parity;
>          semctl( before[ parity], 0, SETVAL, argument);
>        }
>        ...
>
> But when I try to compile it I get the error message
> "CountSharer.c:44: error: storage size of 'argument' isn't known".

You will need either a good book or access to manual pages.  On my
system "man semctl" tells me:

"The calling  program must define this union as follows:

union semun {
int              val;    /* Value for SETVAL */
struct semid_ds *buf;    /* Buffer for IPC_STAT, IPC_SET */
unsigned short  *array;  /* Array for GETALL, SETALL */
struct seminfo  *__buf;  /* Buffer for IPC_INFO
(Linux specific) */
};"

so it is up to you to define this union when you want to use those
features of semctl (those that require a fourth argument).  Your local
system may require something else, but the manual pages should tell you.

--
Ben.

[ Post a follow-up to this message ]

"Ben Bacarisse" <ben.usenet@bsb.me.uk> schrieb im Newsbeitrag
news:87d4wa4703.fsf@bsb.me.uk...
> kvnsmnsn@hotmail.com writes:
>
> You will need either a good book or access to manual pages.  On my
> system "man semctl" tells me:
>
>  "The calling  program must define this union as follows:
>
>     union semun {
>           int              val;    /* Value for SETVAL */
>           struct semid_ds *buf;    /* Buffer for IPC_STAT, IPC_SET */
>           unsigned short  *array;  /* Array for GETALL, SETALL */
>           struct seminfo  *__buf;  /* Buffer for IPC_INFO
>                                       (Linux specific) */
>     };"
>
> so it is up to you to define this union when you want to use those
> features of semctl (those that require a fourth argument).  Your local
> system may require something else, but the manual pages should tell you.
Any idea/reason why that union semun is not part of sys/sem.h?

Bye Jojo

[ Post a follow-up to this message ]

On 2007-09-23 08:58, Joachim Schmitz <nospam.jojo@schmitz-digital.de> wrote:
> "Ben Bacarisse" <ben.usenet@bsb.me.uk> schrieb im Newsbeitrag
> news:87d4wa4703.fsf@bsb.me.uk... 
> Any idea/reason why that union semun is not part of sys/sem.h?

The prototype of the function semctl is

int semctl (int __semid, int __semnum, int __cmd, ...);

I have the nasty suspicion that the union was introduced into the
man-page as a semi-formal way of declaring that the fourth parameter
could be either an int or one of three pointer types but wasn't actually
intended to be used as a union. On most unix systems it doesn't make a
difference whether you call

semctl (sem_id, 0, SETVAL, 15);
or
union semun su;
su.val = 15;
semctl (sem_id, 0, SETVAL, su);

resp.

struct semid_ds ds;
semctl (sem_id, 0, SETVAL, &ds);
or
union semun su;
su.buf = &ds;
semctl (sem_id, 0, SETVAL, su);

either because int and pointers are the same size or because at least
the first 4 parameters are passed in registers.

As further evidence for this suspicion I cite the OSF/1 man-page for
semctl (ca. 1992), which declares the union as:

union semun {
int val;
struct semun *buf;
ushort array[];
};

which isn't even legal C, and if the DEC C compiler supported array[] as
an extension (many compilers at that time did) the semantics would not
have been the same as that of ushort *array (which was obviously
intended).

hp

--
_  | Peter J. Holzer    | I know I'd be respectful of a pirate
|_|_) | Sysadmin WSR       | with an emu on his shoulder.
| |   | hjp@hjp.at         |
__/   | http://www.hjp.at/ |	-- Sam in "Freefall"

[ Post a follow-up to this message ]