From: takach AT ece DOT iit DOT edu (Andres Takach)
Subject: bug report
12 Feb 1997 19:14:29 -0800
Approved: cygnus DOT gnu-win32 AT cygnus DOT com
Distribution: cygnus
Message-ID: <9702121936.AA02046.cygnus.gnu-win32@mar>
Original-To: gnu-win32 AT cygnus DOT com
X-Sun-Charset: US-ASCII
Content-Length: 1351
Original-Sender: owner-gnu-win32 AT cygnus DOT com


There seems to be a problem in the read() (getc etc.) when
reading binary files. A character 1A can confuse the read
to think that it has found the end of file.

This problem also happens when using the gnu-win92 "od" program.
In fact it also happens on the perl implementation for win 95/NT.
The perl problem happens on even NT machines. 

I have included below a simple c program that exposes the problem.

Andres Takach


#include <stdio.h>
#include <stdlib.h>

/* this program illustrates the fact that if the character 1A is
  present in a binary file, getc() will return an EOF at that point
  (same can be said about read() etc.)  The output of this program
  should be 13 (13 chars in file "temp") but instead it is 2.
  Program "od" will also quit early and same can be said about
  perl for windows 95/NT.
*/

main(){
  FILE *fp;
  int ch;
  int count = 0;
  int i;
  fp = fopen("temp","w");

/* Write two random chars to "temp"*/
  putc(255,fp); 
  putc('a',fp);
  
/* Write offending char 26 (Hex 1A) to "temp" */
  putc(26,fp);    

/* Write 10 more 'a's to file "temp"  and close file */
  for(i=0; i < 10; i++) putc('a',fp);  /* 10 more bytes */
  fclose(fp);

/* Attempt to count # of bytes in "temp" */
  fp = fopen("temp","r");
  while( getc(fp)!= EOF) count++;
  /* should be 13 bytes */
  printf("Number of bytes: %d\n",count);
}
-
For help on using this list, send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".