Mailing-List: contact cygwin-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-subscribe AT sources DOT redhat DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin/>
List-Post: <mailto:cygwin AT sources DOT redhat DOT com>
List-Help: <mailto:cygwin-help AT sources DOT redhat DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-owner AT sources DOT redhat DOT com
Delivered-To: mailing list cygwin AT sources DOT redhat DOT com
Message-ID: <1DB8BA4BAC88D3118B2300508B5A552CD9275B@mail.fitlinxx.com>
From: David Bolen <db3l AT fitlinxx DOT com>
To: cygwin AT sources DOT redhat DOT com
Subject: read(0) when stdin is a Windows socket handle returning EACCES
Date: Fri, 14 Jul 2000 03:54:54 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"

I'm running into a strange issue with read() failing to read from stdin when
stdin is a socket connection that I was hoping someone might be able to
offer any suggestions on how to further diagnose.

I have a server application under NT that accepts client connections and
creates a child process, connecting them to that client connection (more or
less an rexec sort of thing).  Fairly typical stuff.  Until now it's been
named pipe based, but I just added an IP (TCP) connection as an option.  The
problem arises when the child process being executed is a Cygwin based
program (the server is not) and the incoming connecting to the server is
over TCP (thus the handle supplied as stdin/stdout for the child is the
socket).  The Cygwin process fails to read properly from stdin, and in fact
generates EACCES errors on the I/O.  This is with v1.1.2 of the DLL under
Windows NT 4.

This small application is enough to generate the problem as target code to
execute:

#include <stdio.h>
#include <errno.h>

main ()
{
   FILE *outfile = fopen("d:/tmp/ctest.out","w");
   char ch;
   int i;

   while (1) {
      i = read(0,&ch,1);
      if (i < 0) {
         fprintf(outfile,"read error %d\n",errno);
         break;
      } else if (i == 0) {
         fprintf(outfile,"read EOF\n");
         break;
      } else {
         fprintf(outfile,"[%02X]",ch);
      }
      fflush(outfile);
   }
}

In my test client, I'm sending the bytes 0x18 0x00 0x00 0x00 to the server.
If that is run with the above code as the child process, with a named pipe
connection, then the output file shows: "[18][00][00][00]read EOF", but if
instead I make an IP connection to the daemon the output file from the above
code shows "[18]read error 13".  It gets the first byte but then gets the
EACCES error.  The above code aborts at that point but if I don't, I'll get
EACCES continuously for every read() after that first.

I haven't been able to get strace to work within the server environment
(e.g., trying to get the server to execute strace which is executing the
test code) and an strace of the above code on its own of course works fine.

Clearly there is some behavior either happening within the cygwin DLL or by
the DLL calling Win32 operations that "burps" after that first I/O when the
stdin handle at process creation is a Windows socket handle, but I'm sort of
stumped as to how to attack this further, so any suggestions would be
appreciated.

-- David

/-----------------------------------------------------------------------\
 \               David Bolen            \   E-mail: db3l AT fitlinxx DOT com  /
  |             FitLinxx, Inc.            \  Phone: (203) 708-5192    |
 /  860 Canal Street, Stamford, CT  06902   \  Fax: (203) 316-5150     \
\-----------------------------------------------------------------------/


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe AT sourceware DOT cygnus DOT com