From: Lyle Johnson <lyle@...>
Date: 2003-05-29T05:17:48+09:00
Subject: [Q] Should I use NUM2UINT or NUM2ULONG?

I'm looking at an FXRuby bug that I suspect has to do with the size 
differences between 32-bit and 64-bit operating systems.

In FOX (the C++ library) a color value is stored as an "FXuint", which 
is typedef'd as a 32-bit unsigned integer. It is *always* this size, 
regardless of the underlying platform's sizes for ints, longs, etc. If I 
am presented with a Ruby Integer (which can be either a Fixnum or a 
Bignum) I want to convert that number to an FXuint.

So far, I've been using Ruby's NUM2UINT macro for this purpose, i.e.

     FXColor color = NUM2UINT(value);

and this works fine on my Linux and Windows boxen. But one of my users 
is trying the code out on a DEC box, which I'm pretty sure is a 64-bit 
machine, and he's getting an error when such conversions are attempted, e.g.

     bounce.rb:105:in `setForeground': integer 4294967295 too big to 
convert to `int' (RangeError)

This value should fit into an unsigned integer (it is equal to UINT_MAX) 
but the error message indicates that it's trying to convert it to an 
'int' instead. I'm still waiting on some additional information from 
this user, but I'm guessing that SIZEOF_INT is less than SIZEOF_LONG in 
"config.h", and so the NUM2UINT gets expanded to NUM2INT. And the value 
(4294967295) is too big to fix into a signed integer, and so an 
exception is raised.

So now I'm wondering, should be using NUM2ULONG for this conversion 
instead? I think it's probably a bug that NUM2UINT gets this wrong, but 
at this point I just want something that will work correctly.

Thanks for any suggestions,

Lyle