This code does not work on 32 bit big endian machines:

I think it just needs to be changed to an int. But I wonder if we have other code like that that uses a long instead of an int. Maybe all that code should use that macro.

gcc also has built in macro's for that, I suggest we use that by default instead.