X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
From: Radical NetSurfer <RadSurfer AT yahoo DOT com>
Newsgroups: comp.os.msdos.djgpp
Subject: MAJOR PAIN: Reading Integers!
Date: Sat, 25 Dec 2004 02:11:42 -0500
Message-ID: <fi3qs05k7quvrsordlba97rkmdkvrputca@4ax.com>
X-Newsreader: Forte Agent 1.93/32.576 English (American)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Organization: Velocity.Net
Cache-Post-Path: web.velocity.net!unknown AT 66-211-206-248 DOT velocity DOT net
X-Cache: nntpcache 3.0.1 (see http://www.nntpcache.org/)
Lines: 126
X-Complaints-To: abuse AT newshosting DOT com
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Here's a little something I hope triggers some discussion...

DJGPP has *MANY* features simply not found in most compilers,
Borland, visualc, etc...
but even DJGPP still requires a bit of thoughtful ingenuity to
get access to some values in a file....

Case in point is reading of a "Raw Binary Value" directly out of
a file.
By "Raw Binary Value" we mean *NOT* being switched around
by Intel's storage format (big vs little endian).

1. Lets create a binary file, containing an unsigned long:

   outfile = fopen ("bf.bin", "wb");
   fprintf(outfile,"\x12\x34\x56\x78");
   fclose(outfile);
   //4-bytes expected and NOTHING else.

2. Object here is to now obtain in an unsigned long,
    with as little effort as possible the ORIGINAL
    ALIGNMNET of BYTES (defeating intel format);
    what I am calling "raw binary value"

3. Attemp #1:

 struct lbf {
  unsigned long bf:32;
 } abf;

void main(void) {
FILE *infile;
FILE *outfile;

   outfile = fopen ("bf.bin", "wb");
   fprintf(outfile,"\x12\x34\x56\x78");
   fclose(outfile);

   infile = fopen ("bf.bin", "rb");
   fread(&abf, sizeof(abf), 1, infile);
   fclose(infile);

  printf("bf = %lu = %X\n", abf.bf, abf.bf );

  /* Guess what! Its *NOT* a raw bitfield!!
      It's actually in Intels format!  */
};

4. Attempt #2:

unsigned long abf;
unsigned char Buf[4];

void main(void) {
FILE *infile;
FILE *outfile;

   outfile = fopen ("bf.bin", "wb");
   fprintf(outfile,"\x12\x34\x56\x78");
   fclose(outfile);

   infile = fopen ("bf.bin", "rb");
   fread(Buf, sizeof(Buf), 1, infile);
   fclose(infile);

   abf = (unsigned long)((Buf[0] << 24) | (Buf[1] << 16) |  (Buf[2] <<
8) | Buf[3]);

  printf("bf = %lu = %X\n", abf.bf, abf.bf );

  /* Works at the expense of Shifts & Ors, Casting required. */

} //main

WHICH ENDS up being resolved into:

5. Attempt #3:

unsigned long abf;
unsigned char Buf[4];

void main(void) {
FILE *infile;
FILE *outfile;

   outfile = fopen ("bf.bin", "wb");
   fprintf(outfile,"\x12\x34\x56\x78");
   fclose(outfile);

   infile = fopen ("bf.bin", "rb");
   fread(Buf[3], sizeof(char), 1, infile);
   fread(Buf[2], sizeof(char), 1, infile);
   fread(Buf[1], sizeof(char), 1, infile);
   fread(Buf[0], sizeof(char), 1, infile);
   fclose(infile);

   abf = cvul( Buf );  //Expect Unsigned Char Address Space to be UInt

  printf("bf = %lu = %X\n", abf.bf, abf.bf );
 
  /*  Eureka! Success !!
      At the expense of using a String-Array,
                                         ... *4* separate fread calls!
      We still have to resolve the value as an unsigned long, which
      is accomplished thru trivial pointer/casting of the contents of
      the unsigned char buffer space...
  */

} //main

6. Attempt 6,7,8,9,etc.
    Are just too freaky to bother posting!

7. Conclusion:
    Except for being "clever" at using *4* fread's
    or by re-arranging a unsigned char, byte-by-byte
    using shifts and ors,
    I just see NO WAY to directly read in Integer
    Values without being "corrupted" by Intel's expectations!

    what other method(s) are available to achieves this that
    I may of missed?


ANY COMMENTS WELCOMED.  Thanks.