Re: [AD] replace xlock/xunlock with synchronized{}

[Jon Rafkind <workmin@xxxxxxxxxx>]

Peter Wang wrote:

> On 2006-04-05, Jon Rafkind <workmin@xxxxxxxxxx> wrote:
>  
>
> > Would anyone be opposed to replacing all the XLOCK()/XUNLOCK() blocks in 
> > the x11 code of Allegro with a #define'd syntax that does the same 
> > thing, with preferably the same name as java's sychnronization keyword 
> > 'synchronized' ?
> >
> > Basically do this
> >
> > #define synchronized for ( XLOCK(); XUNLOCK(); /**/ )
> >    
>
> Clever.
>  
Actually the a real working define needs a temporary variable so that 
the body of the loop gets run once. Im not sure how compliant this is, 
but it works with gcc 3.4.3:

#define synchronized for( int __l( ! XLOCK() ); __l; __l = 0, XUNLOCK() ) )

All that would be required is if XLOCK() returned a value, 0 for success 
and 1 for failure.

>  
>
> > and then instead of
> >
> > int _xwin_create_window(void)
> > {
> >  int result;
> >  XLOCK();
> >  result = (*_xwin_window_creator)();
> >  XUNLOCK();
> >  return result;
> > }
> >
> > would be
> >
> > int _xwin_create_window(void)
> > {
> >  int result;
> >  synchronized{
> >      result = (*_xwin_window_creator)();
> >   }
> >  return result;
> > }
> >
> > Yes I use Java a lot these days and I find this very clean. Plus it 
> > would prevent the mistake of forgetting to use XUNLOCK() occasionally.
> >    
>
> Unfortunately there is the risk of incorrect but seemingly correct code
> like this:
>
>    synchronized {
> 	if (foo) {
> 	    return 1;
> 	}
>    }
>
>  
Well.. yea I suppose that could happen. Im willing to let it slide if 
other people can too.