From: shoham AT enel DOT ucalgary DOT ca (Idan Shoham)
Subject: Re: djgpp and the 386SX
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Date: Sun, 9 Oct 1994 15:19:30 -0600 (MDT)

... stuff deleted for brevity ...

> This is simply not true!  A TRUE 32 bit compiler is 32-bits throughout, 
> instead of using 16 or 32 bits here and there.  This is a major 
> complaint of users of Borland and Microsoft compilers that are touted to 
> be 32-bit compilers where they don't use 32-bits for their integers.  

Yup.  very very true.  Borland==Microjunk==junk.  8)

... stuff deleted for brevity ...

> As far as Unix programs and such are concerned, what's wrong with 
> assuming that the size of an int is the same as the size of a pointer?  

Well, for one, it's not necessarily true.  For instance, on my DEC Alpha,
sizeof(int)=4, sizeof(int *)=8

> This is almost a given in the C programming world, or at least is should 
> be.  I conceed that assuming the length of any pointer is bad programming 
> practice -- and, in general, assuming the byte ordering of any "pointer" 
> or "int" type is bad programming practice; but for those programs (device 
> drivers / kernels) that require a low-level understanding of the 
> architecure[sp] that it is running on, what else are the programs to 
> assume?  There is a certain level of non-portability in every program, it 
> just depends on what the program is designed to accomplish as to how 
> non-portable it is.

They should assume that a pointer is a pointer, and its size is "pointer size".
You just can't make these assumptions and expect your programs to work
on any machine....  Too many people grew up on just 1 or 2 platforms, and
generalize bad assumptions, and I've had too many times where I've had to
fix their stuff on my systems.  arrgh!  ;)

... more stuff deleted for brevity ...

Ah well, just my 2c worth'..

-- Idan