Subj : Re: trying to make own version of pthread_exit()
To   : comp.programming.threads
From : steve
Date : Tue Mar 22 2005 09:36 pm

In article <d1pa00$dk1$1@news.net.uni-c.dk>, JS <sfsdfs@asdas.com> wrote:
>I am trying to make my own implementation of pthread_exit(). I have read:
>
>http://www.llnl.gov/computing/tutorials/workshops/workshop/pthreads/man/pthread_exit.html
>
>To prevent confusion I call the funtions mythread instead of pthread.
>
>
>When a thread calls mythread_exit() it should be terminated safely. But it
>should also store a "termination status void pointer" that any joining
>threads can access.

Yes, and the value stored needs to be valid after the thread stack
has been deleted.

>When I create a new thread, start_function is called which then calls
>mythread_exit with the pointer that it returns. Is this the right use of
>pthread_exit function? I store the return value that start_function produces
>in a void pointer k.

The value to be returned can either be something int-like cast into a void *,
or a pointer to something in the heap.  A pointer to something on the
stack won't work, because the stack will be going away.

[ snip ]

>void mythread_exit(void *retval){
> void * k;
> k = retval;
>
> printf("%d",k);
>
>}

Uh, that doesn't quite do enough if you want to be like pthread_exit().
pthread_exit() can be called from anywhere in the execution flow and
will stop that execution flow.

>void *start_function(void *arg){
>    int x = 22;
>    int *ip;
>    ip = &x;
>
> mythread_exit(ip);

If you're expecting to see something related to 22, you would be better
off saying "mythread_exit((void *)22);".  In this case, x goes out of
scope before the return value is fetched.  And your implementation of
mythread_exit() above is printing the value of the pointer.

>return ip;
>}

The return is redundant.

If you're serious about implementing a thread environment, I would
suggest thinking very carefully about the lifetime of various objects;
it's nasty to debug stack slams in a multithreaded environment, and
debugger support is difficult to write, too.
-- 
Steve Watt KD6GGD  PP-ASEL-IA          ICBM: 121W 56' 57.8" / 37N 20' 14.9"
 Internet: steve @ Watt.COM                         Whois: SW32
   Free time?  There's no such thing.  It just comes in varying prices...

.