Re: [Boost-Users] Re: weak_ptr

From: Peter Dimov (pdimov_at_[hidden])
Date: 2003-07-23 05:50:09

From: "Jason Winnebeck" <yg-boost-users_at_[hidden]>
[...]
> >   //waitForPlayer returns the connected player, or NULL if the
connection
> >   //process was aborted by the user.
> >   PongClient::sptr waitForPlayer() {
> >     LockCV lock(sync);
> >
> >     while (!(player.lock()) && !kbhit()) {
> >       //We wait for 250ms to recheck kbhit for pressed keys.
> >       sync.timedWait(250);
> >     }
> >     if (!(player.lock())) {
> >       //We were woken up by a keypress, so refuse any further
connections.
> >       accept = false;
> >
> >       //We don't need to wait around if anyone is in the middle of
connecting,
> >       //because shutting down GNE will close any open connections, and
we will
> >       //be closing down if we aborted this connection.
> >
> >       //Return that no client connected.
> >       return PongClient::sptr();
> >     }
> >
> >     return player.lock();
> >   }

Unfortunately I don't have the time to properly review your code. I'll just
rewrite the above in a more "idiomatic" way (I hope I'm not missing
something.)

PongClient::sptr waitForPlayer()
{
    LockCV lock(sync);

    for(;;)
    {
        if( PongClient::sptr p = player.lock() )
        {
            return p;
        }
        else if( kbhit() )
        {
            accept = false;
            return PongClient::sptr();
        }
        else
        {
            sync.timedWait(250);
        }
    }
}

Even more "idiomatic" would be to use shared_ptr<...> instead of the ::sptr
typedef. Hiding the type of the smart pointer behind a typedef usually
doesn't help much and can lead to subtle errors since the semantics of the
various smart pointers differ in some corner cases.

HTH