perl -pn -i.bak -e 's/\032//g' filename
[download]

If by EOF you mean the End Of File marker then depending on the OS this may or may not work:

   perl -pi -e's/[\cZ\cD]//g' file
[download]

/J\

while (<FILE>) {
[download]

You can't create a file that has no EOF at all. Not even on windows. You can create a file that appears very long (like /dev/zero on unix), but you have to work at it. Don't worry, whatever file you try to read on your system, the system will let you know when you've reached the end of file.

If you cat 2 files together, only one EOF is placed at the end of the combined files, automatically.

You will never find a normal file for while the while ! eof construct is an infinite loop. Maybe you should explain your entire situation. Are you actually going to be dealing with normal files, or are these sockets or some other kind of pseudofile?

// DeEOF.cpp : Reads in a file and outputs the contents, omitting
// all EOF characters.
//
//  Parameter:   <inputfile> - the file containing the undesired EOFs.
//
//
// Written by Brendan Cunnie 
//

#include "stdafx.h"
#include <string.h>
#include <fstream.h>
#include <iostream.h>
#include <ios.h>
#define ASCII_EOF 26
 
int main(int argc, char* argv[])
{
        bool    fDisplayHelp = false;
        if (argc == 1) fDisplayHelp = true;
        if ( (argc>1) &&
                 ( (strcmp(argv[1],"-?")==0) || (strcmp(argv[1],"-h")=
+=0)) ) {
                fDisplayHelp = true;
        }
        if (fDisplayHelp) {
                        cout << "Takes a file and outputs it, strippin
+g all EOF characters." << endl;
                        cout << "Usage:" << endl;
                        cout << "  DeEOF <inputfile>" << endl;
                        return (0);
        }
 
        ifstream ifile ( argv[1], ifstream::binary);
        char    ch;
        for (ch = ifile.get(); !ifile.eof(); ch = ifile.get()) {
                if (ch != ASCII_EOF) {
                        cout << ch << flush;
                }
        }
        return 0;
}
[download]

Edited by Chady -- code tags

How about if you update this solution by putting "<code>" at the beginning of the cpp code, and putting "</code>" at the end of it, so the rest of us can read it more easily.

IIRC, the "^Z" character (0x1A in hex, \032 in octal, 26. decimal) has always been used -- and is still used -- by MS systems as the byte value that marks the end of a text file. That is, if a file is being opened and read in text mode on an MS system, then there will be an EOF condition when a ^Z byte is encountered.

Of course, if the file is opened and read in binary mode, then ^Z has no special meaning, and will be treated the same as every other possible byte value. This is important, since many non-text files (containing image, audio, compressed, compiled executable or similar kinds of data) tend to contain bytes whose values happen to be 26. (i.e. 0x1a, 032, ^Z), and using text mode on such files will cause a premature EOF condition -- not good. (There are other evils that arise when treating non-text files with MS text-mode i/o, but I shouldn't digress...)

As for removing ^Z from a file, well... Obviously, if you do this globally on a non-text file, this is simply a form of data corruption -- whatever the original data may have been, it will be garbage after all the ^Z's are removed.

If, using an MS system, you want to do this on a real DOS/Windows text file (where there is just one ^Z, at the very end), I believe you would have to open both input and output files in "binary mode"; if you read such a file in text mode (like you're "supposed to"), the program would never see the ^Z -- the OS intercepts it on reading and appends it on writing, and the program handling files in text mode never sees this character. You can only read and write ^Z explicitly in your program when handling files in binary mode. (That's the main and traditional use of perl's "binmode" function, though now as of Perl 5.8, this function extends to cover other things as well, like character encoding.)

I was intrigued by the question and had a go at trying to do it. I often find myself using a hex dump utility to 'see what's actually there'. Looking in Perl would be useful. I've not had experience with binary mode before so it was about time I did.
(dummy.txt has 99 'a's and a 'b')

use strict;
use warnings;
use Fcntl;

my $stream;

# 'or die ...' removed for clarity

sysopen(DUMMY, "dummy.txt", O_RDWR | O_BINARY);
my $bytes_read = read DUMMY, $stream, 128;

for ( my $i=0; $i<= $bytes_read; $i++ ){
  my $char = substr( $stream, $i, 1 );
  print $i, ": ", ord( $char ), " => *", $char, "*\n";
}
[download]

produces...

0: 97 => *a*
1: 97 => *a*
2: 97 => *a*
3: 97 => *a*
... etc
98: 97 => *a*
99: 98 => *b*
100: 0 => **
[download]

Nowhere near it (not even new lines). I found the docs quite intimidating, clearly the cross platform issues are tricky. (I was reading one article that mentioned CP/M!)
Any pointers?
activestate 5.8 on winXP

Well that's cleared that up then. I was just starting to get interested!
update: Sorry, looks better with code tags