Experimental Physics and
Industrial Control System

John William Sinclair <[email protected]> · Tue, 31 Oct 2006 16:35:08 -0500

Of course I'm not suggesting (and I should have made this clear - sorry)
that this is a fix. This is just data to show that making NULL correct
for this architecture fixes this single instance of the problem.

John

On Tue, 2006-10-31 at 15:26 -0600, Andrew Johnson wrote:
> >> Putting the following define in all .c source files (after the last
> >> #include)
> >>
> >>   #define NULL 0L
> >>
> >> does indeed fix things.
> 
> It may fix things on your 64-bit machine, but it won't everywhere -- 
> you're providing a long argument when the XtVaSetValues() routine is 
> looking for a pointer; the two may still be different sizes on other 
> machines.
> 
> NULL is a strange beast.  On my linux-x86 machine gcc 4.0.2 defines it 
> as ((void *)0) whereas g++ 4.0.2 defines it as __null.  The Solaris C 
> C++ compiler defines it to be 0 in all circumstances; the C compiler 
> uses 0L if we're compiling for LP64 archs, else 0.
> 
> I think that the correct solution is to accept the compiler's definition 
> and just use NULL.  Thus if removing NULL from dbDefs.h solves this 
> problem then that's the right solution for now and it'll be that way 
> from R3.14.9 onwards.
> 
> - Andrew

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System