|
Threads-Devel : |
From: rick68 (rick68_at_[hidden])
Date: 2007-07-02 07:49:55
On 7/2/07, Anthony Williams <anthony_w.geo_at_[hidden]> wrote:
>
> rick68 <rick68_at_[hidden]> writes:
>
> > I try timed_mutex whit linux, but it not work.
>
> > void timed_mutex::do_lock()
> > {
> > int res = 0;
> > res = pthread_mutex_lock(&m_mutex);
> > assert(res == 0);
> >
> > while (m_locked)
> > {
> > res = pthread_cond_wait(&m_condition, &m_mutex);
> > assert(res == 0);
> > }
> >
> > assert(!m_locked);
> > m_locked = true;
> >
> > res = pthread_mutex_unlock(&m_mutex);
> > assert(res == 0);
> > }
> > All while-loop never running, because of
> > boost::timed_mutex::scoped_timed_lock will check m_locked is false.
>
> Yes. The loops are there so that if the mutex is already locked (m_locked
> is
> true), do_lock will wait for the lock to become available. If the mutex is
> unlocked (m_locked is false) then do_lock doesn't have to wait, so doesn't
> run
> the while loop.
If I use boost::timed_mutex::scoped_timed_lock to do lock (m_locked is true)
==<boost/thread/detail/lock.hpp>==
void boost::detail::thread::scoped_try_lock<boost::timed_mutex>::lock()
{
if (m_locked) throw lock_error();
lock_ops<TimedMutex>::lock(m_mutex);
m_locked = true;
}
========================
lock will throw lock_error, so I thought maybe to use while-loop will be
fine.
> So, I change while-loop to do-while-loop:
> >
> > ============================================
> > void timed_mutex::do_lock()
> > {
> > int res = 0;
> > res = pthread_mutex_lock(&m_mutex);
> > assert(res == 0);
> >
> > do
> > {
> > res = pthread_cond_wait(&m_condition, &m_mutex);
> > assert(res == 0);
> > } while (res);
> >
> > assert(!m_locked);
> > m_locked = true;
> >
> > res = pthread_mutex_unlock(&m_mutex);
> > assert(res == 0);
> > }
>
> This will deadlock! If there is no other thread holding the mutex
> (m_locked is
> false) then this will enter the condition wait, but no thread will notify
> the
> condition: deadlock.
Oh, you are right! :p
But I don't have any idea let timed_mutex to wait... :(
> It's work very well~ :)
>
> What exactly are you trying to achieve? If you let us know what you want
> to do
> (in your code), then someone might be able to propose a way of achieving
> that
> without breaking timed_mutex.
I just read about thread code and test it.
Can use condition and mutex to substitute for timed_mutex, but I want to
know how to use timed_mutex.
Good Lucky~
rick