[boost] Re: Lock Classes: Does anyone care.

From: Alexander Terekhov (terekhov_at_[hidden])
Date: 2003-02-20 11:11:59

Kevin Atkinson wrote:
> 
> On Thu, 20 Feb 2003, Alexander Terekhov wrote:
> 
> > > I have changed the definition to:
> > >
> > > #ifdef FAST_MUTEX_INIT_DESTROY
> >                          ^^^^^^^
> >
> > >   static const pthread_mutex_t MUTEX_INIT = PTHREAD_MUTEX_INITIALIZER;
> >
> > Uhmm. What does your "fast destruction" do? Well, looking at the code
> > you've posted, it does nothing... and that's the "fastest" way to leak.
> 
> It depends on the implementation.  On some, including linux, there is
> nothing to free.  

But that might change overnight [on the next start, I mean]... and even 
without any recompiles (rebuilds) on the part of your clients, Kevin.

<illustration source=nptl-0.24/nptl/old_pthread_cond_wait.c>

/* Copyright (C) 2002, 2003 Free Software Foundation, Inc.
   This file is part of the GNU C Library.
   Contributed by Ulrich Drepper <drepper_at_[hidden]>, 2002.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, write to the Free
   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
   02111-1307 USA.  */

#include <errno.h>
#include <stdlib.h>
#include "pthreadP.h"
#include <atomic.h>
#include <shlib-compat.h>

#if SHLIB_COMPAT(libpthread, GLIBC_2_0, GLIBC_2_3_2)
int
__pthread_cond_wait_2_0 (cond, mutex)
     pthread_cond_2_0_t *cond;
     pthread_mutex_t *mutex;
{
  if (cond->cond == NULL)
    {
      pthread_cond_t *newcond;

#if LLL_MUTEX_LOCK_INITIALIZER == 0
      newcond = (pthread_cond_t *) calloc (sizeof (pthread_cond_t), 1);
      if (newcond == NULL)
        return ENOMEM;
#else
      newcond = (pthread_cond_t *) malloc (sizeof (pthread_cond_t));
      if (newcond == NULL)
        return ENOMEM;

      /* Initialize the condvar.  */
      (void) pthread_cond_init (newcond, NULL);
#endif

      if (atomic_compare_and_exchange_acq (&cond->cond, newcond, NULL) != 0)
        /* Somebody else just initialized the condvar.  */
        free (newcond);
    }

  return __pthread_cond_wait (cond->cond, mutex);
}
compat_symbol (libpthread, __pthread_cond_wait_2_0, pthread_cond_wait,
               GLIBC_2_0);
#endif

</illustration>

This is from the latest NPTL code drop. BTW, it seems that it's broken 
with respect to memory visibility on MPs with relaxed memory ordering, 
but never mind. ;-)

regards,
alexander.