Re: NSStream update

[Richard Frith-Macdonald]

On 21 Feb 2006, at 19:18, Derek Zhou wrote:

> "Andrew Ruder" <aeruder@gmail.com> writes:
>
> > > 2. make the ivar/setter generic somehow ... eg by using a void*
> > > rather than an int.  Actually a windows handle is an unsigned long
> > > thought I'm not sure it's guaranteed to stay that way ... using  
> > > void*
> > > in the generic part should be safe though.
> >
> > The worst thing about the void * is that the number of warnings
> > generated on a amd64 machine is pretty staggering on NSStream.  I've
> > included a sample output below.  Can't this be done with typedef's?
> >
> > - Andy
> I don't have a 64bit machine so I cannot verify, but In
> Foundation/NSRunloop.h there is:
>
> - (void) addEvent: (void*)data
>              type: (RunLoopEventType)type
>           watcher: (id<RunLoopEvents>)watcher
>           forMode: (NSString*)mode;
>
> Where data is the file descriptor (int) in Unix. So how does that
> work?

You call it as [loop addEvent (void*)(intptr_t)desc type: type  
watcher: watcher forMode: mode]

The double cast tells the compiler how to pass it and not to generate  
a warning.

The (intptr_t) tells it to extend an integer value to the size of a  
pointer.
The (void*) tells it that it is OK to treat the value as an abstract  
pointer.

PS. I made sure the code in svn does not generate warnings.