write and fwrite send extra bytes when writing binary data to stdout

[Emilio Hernandez-Garcia]

   Hi all,
This code is supposed to use 'write' to write in binary format a float 
array of 12
elements (thus 4*12=48 bytes) to stdout (on execution it is redirected 
to a file):

/* ************************************* */
#include <stdio.h>
main()  {
  float const vector[12]={1.,0.2,2.,0.,1.,-1.,512.,512.,0.01,100.,1.,0.};
  fprintf(stderr," %ld bytes writen to stdout\n",write(1,vector,48));
}
/* *************************************** */

- I compile it with gcc 3.2-3 under cygwin 1.3.17-1.
- The fprintf command tell us that the program has written to stdout 48 
bytes.  But looking
at the actual file size, it is of 49 bytes.
- Examining the file with an hex viewer, there is an extra byte 0D
between the 8th and the 9th element of the array. This makes the file
useless to be used by external applications.
- The extra byte does not appear if you change some of the numbers (for 
example replace 0.01 by
0.001) but if you try to write a large enough array, extra bytes always 
appear somewhere.
- The same behavior appears by using fwrite(...,stdout). It does not
happen if fwrite writes to a file open with fopen. I've tried the same 
code under gcc and
linux Suse7.*, several versions of c compilers under SGI and HP unix, 
... ,and the extra bytes never appear.

if you try to read the file generated under cygwin with read (or 
fread(...,stdin)) the behavior is still
more curious: the following code reads 'vector' from stdin (which on
execution is conected to the binary file written before):

/* ************************************** */
#include <stdio.h>
main()  {
        float vector[12];
        int i;
        fprintf(stderr,"%ld bytes read\n",read(0,vector,48));
        for( i=0; i<12; i++){fprintf(stderr,"%f ", vector[i]);}
}
/* ************************************** */

The fprintf statement tell us that the program has read 47 bytes, which
reflects some error. But in fact it has read 49, since all the numbers
in the file have been read correctly.

Still more: if you read stdin from a file written by the first program under
linux, where it behaves correctly and outputs 48 bytes, the reading 
program also
reads in the numbers correctly, and reports the correct byte count.

All of this is quite puzzling, and frustrating when trying to use cygwin
to write binary to be piped to other applications.  Any insight?

I attach the output of 'cygcheck -s -v -r > cygcheck.out'
-- 
                Emilio Hernandez-Garcia

Instituto Mediterraneo de Estudios Avanzados (IMEDEA) Tel:34 971 171307
CSIC-UIB, Campus Universitat Illes Balears            Fax:34 971 172309
E-07122 Palma de Mallorca                  e-mail: emilio@imedea.uib.es
Spain                                 http://www.imedea.uib.es/~emilio/


-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: cygcheck.out
URL: <http://cygwin.com/pipermail/cygwin/attachments/20030225/adc62781/attachment.ksh>
-------------- next part --------------
--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/