Subj : Re: pthread_cancel
To   : comp.programming.threads
From : doug
Date : Mon May 09 2005 09:08 am

"Greg Law" <law@privacy.net> wrote in message 
news:2l81pnF9h89tU1@uni-berlin.de...
> Hi,
>
> I am calling pthread_cancel on a thread that has canceling enabled, of 
> deferred type.  The thread I am canceling never makes any calls to 
> testcancel, or waits on any condition variables.  So it surprises me that 
> the cancelation has any effect at all.  That is, in the following program 
> the thread I've created is terminated (to my surprise).  I guess I've not 
> understood correctly when threads may be canceled?
>
> Cheers,
>
> Greg
>
>
> #include <pthread.h>
> #include <assert.h>
>
> void
> Start (void *a) {
>    int tmp;
>    int r = pthread_setcancelstate (PTHREAD_CANCEL_ENABLE, &tmp);
>    assert (!r);
>    r = pthread_setcanceltype (PTHREAD_CANCEL_DEFERRED, &tmp);
>    assert (!r);
>
>    while (1) {
>       printf ("%d\n", tmp++);
>    }
> }
>
> int
> main (void) {
>    int r = 0;
>    pthread_t p;
>
>    r = pthread_create (&p, NULL, (void*)Start, 0);
>    printf ("%d: %ld\n", r, p);
>
>    r = pthread_cancel (p);
>    printf ("back from cancel (%d)\n", r);
>    while (1);
> }

If I remember correctly, I think POSIX states that certain system calls 
(read, write, etc.) are cancellation points.  So the printf() in your 
Start() routine would be a cancellation point.
What version of OS & threading lib are you using?  I know that until 
recently, LinuxThreads hand't actually implemented this, though (e.g. see 
manpage for pthread_cancel on an older Fedora Core 3 system).  It may have 
been fixed recently, or may be correct in NPTL.

.