From: Bill Godfrey <bill-godfrey AT usa DOT net>
Newsgroups: comp.lang.c,comp.os.msdos.djgpp
Subject: Re: typedef'ing a 24-bit integer
Date: Wed, 26 May 99 22:52:29 
Message-ID: <19990526.68595010.14752@bacchae.demon.co.uk.228.60.172>
References: <37459ABF DOT FF51CA41 AT aaual DOT ualg DOT pt> <927396955snz AT genesis DOT demon DOT co DOT uk>
NNTP-Posting-Host: bacchae.demon.co.uk
X-NNTP-Posting-Host: bacchae.demon.co.uk:212.228.60.172
X-Trace: news.demon.co.uk 927759119 nnrp-02:2887 NO-IDENT bacchae.demon.co.uk:212.228.60.172
X-Complaints-To: abuse AT demon DOT net
Lines: 19
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

fred AT genesis DOT demon DOT co DOT uk (Lawrence Kirby) writes:

> No type in C is guaranteed to occupy a specific amount of space in memory.
> What C guarantees is minimum ranges for specific types. Bit-fields
> act like specific width types but there are few guarantees about their
> storage. Why do you need this anyway? Such a requirement usually
> indicates a poor approach.

(I can't speak for the original questioner, but...)
Maybe to store a large array of integers each of which only requires
24 bits of range.

To do this, I'd probably keep a malloc'd array of unsigned characters, and
write a macro to take an index, extract [i*3] [i*3+1] and [i*3+2] and
shove them into a long using << and | operators.

What to do when CHAR_BIT != 8 is left as an excercise to the reader.

Bill, bit flippingly.