NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/lang/c/17084

Internet Message Format | 1992-11-24 | 2.6 KB

Path: sparky!uunet!oracle!unrepliable!bounce
Newsgroups: comp.lang.c
From: [e-mail] (William Kaufman)
Subject: Re: Typedefing builtin... thanks.
Message-ID: <[e-mail]>
Sender: [e-mail] (Oracle News Poster)
Nntp-Posting-Host: hqsun2.us.oracle.com
Organization: Oracle Corporation, Redwood Shores CA
References: <[e-mail]> <[e-mail]> <[e-mail]>
Date: Tue, 24 Nov 1992 02:57:30 GMT
X-Disclaimer: This message was written by an unauthenticated user
              at Oracle Corporation.  The opinions expressed are those
              of the user and not necessarily those of Oracle.
Lines: 40

In article <[e-mail]> [e-mail] (Ed Driscoll) writes:
] In article <[e-mail]> [e-mail] (Ray Peck) writes:
] >[...#define'ing WORD...]
]
] I've been thinking about doing it that way myself, but it seems to me that
] you would want two variants of each size: one for when you need EXACLTY
] that number of bits, and one when you just need AT LEAST that number of
] bits.

    If you ever need *exactly* n bits, you'll find yourself frustrated
eventually.  What set of #defines will give you exactly 8 bits on a
machine with 9-bit chars and 36-bit ints?

    (Before you reply: Yes, they exist.  Honeywell/Bull machines are
like this.  Also, some machines allowed 6-bit bytes, at the programmers
option.  And, saying you'll never work on one of these is the surest way
to make sure that you will.)

    Using a set of "at least" macros, though, is fine, as long as you're
willing to limit yourself to reasonably large machines: there *are*
machines out there without 32-bit ints, after all.

] The former type would be used where the exact size matters (say,
] something you're reading from a file, or something used for bit flags).

    You should never write binary data when you expect to move data
between different machines; and, if you aren't moving binary data
between machines, you'll never need this type.  (See the FAQ, section
17.2.)

    The suggestion wrt flags is correct, though; and, you can restrict
the set of macros to unsigned types.  But, because you can't be certain
about the size of the types available, this is really an "at least"
type: you need to address at least N bits, not exactly N bits.

                                           -- Bill K.

Bill Kaufman,          | "Hush my darling / Be still my darling /
Corporate Lackey       |  The lion's on the phone,..."
[e-mail] |              -- "The Guitar", They Might Be Giants