On 06/18/2015 10:35 AM, Malahal Naineni wrote: > Soumya Koduri [skod...@redhat.com] wrote: >> Right. But even on success, "retval" need not be set until the >> thread exits with a value other than NULL. >> >> From the manpage, >>>>> >> If retval is not NULL, then pthread_join() copies the exit >> status of the target thread (i.e., the value that the target thread >> supplied to pthread_exit(3)) into the location pointed to by >> *retval. If the target thread was canceled, then PTHREAD_CANCELED is >> placed in *retval. >> <<<<< > > The retval that the man page refers there is actually passed in arg to > pthread_join. This is never null in your case as you pass &retval. > >> So to check if the retval is not NULL or if thread in fact had exit >> with a value, had declared it as pointer. > > You don't need to declare it as a pointer if your thread exits with an > integer. > >> No at present, we do not see these messages as UP_THREAD returns >> NULL on exit. > > The only way the existing code makes sense is if your upcall thread > exits with a pointer. An integer exit should cause the code crash. Are > you really intending to return a valid pointer from your upcall thread > rather than an integer? > I thought pthread_exit() always returns a pointer which gets assigned to retval of pthread_join(). I assume this is the flow -->
pthread_join (thread, void **retval_join) pthread_exit (void *retval_exit) On exit, *retval_join = retval_exit; I tried a sample program as per your suggestion. int ret1; pthread_join (thread, (void **)&ret1); int ret2 = 1; pthread_exit (&ret2); After pthread_join, ret1 has address of ret2 instead of its value. i.e, this led to ret1 = &ret2; Please correct me if I am missing something. Thanks, Soumya > Regards, Malahal. > ------------------------------------------------------------------------------ _______________________________________________ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
