Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!rex!uflorida!gatech!udel!mmdf
From: burgess%creek.decnet@hqhsd.brooks.af.mil (CREEK::BURGESS)
Newsgroups: comp.os.minix
Subject: Re: use only short and long (not int) in struct message
Message-ID: <41145@nigel.ee.udel.edu>
Date: 9 Jan 91 02:37:54 GMT
Sender: mmdf@ee.udel.edu
Lines: 28

Klamer Schutte -- Universiteit Twente <klamer@MI.ELTN.UTWENTE.NL> writes:
>
>HBO043%DJUKFA11.BITNET@cunyvm.cuny.edu (Christoph van Wuellen) writes:
>
>>If struct messages contains only short and long (no 'int'),
>>various parts of the system (user code, FS, MM, Kernel etc) can
>>communicate irrespective of wether they are compiled in 16-bit or
>>32-bit mode.
>
>Why include the assumption that short == 16 bits and long == 32 bits?
>Please use something like int16 and int32 when you mean that.
>This will save you when using a machine where short=16, int=32 and long=64 bits
>
>These names should be typedef'ed somewhere.
>
>Comments?
  Wasn't the original definition of "short" and "long" from K&R explicit?  I
recall a page in the book describing the various lengths of the the types and
"int" being the only machine-specific one.  I think the definition was
short: 8 bits and long: [16||32) i can't remember exactly)] bits.  Besides,
the compilers in question are virtually all written by and maintained by
people on the net.  If short is always eight bits and long is always 32 bits,
then the idea (IMHO) is an excellent method to overcome the problems with
variant "int" sizes.

TSgt Dave Burgess
Armstrong Labs Det 4
Brooks AFB, TX