Sender: nate AT cartsys DOT com
Message-ID: <35AAF20E.FA3193F9@cartsys.com>
Date: Mon, 13 Jul 1998 22:52:14 -0700
From: Nate Eldredge <nate AT cartsys DOT com>
MIME-Version: 1.0
To: "John M. Aldrich" <fighteer AT cs DOT com>
CC: djgpp AT delorie DOT com, "David P. Hack" <hack AT sky DOT net>
Subject: Re: Buffered input
References: <6_Ap1.22$s7 DOT 1149262 AT alpha DOT sky DOT net> <6o79as$913$1 AT slave1 DOT aa DOT net> <35A7E6DA DOT EA2C4522 AT cs DOT com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

John M. Aldrich wrote:
> 
> Roger W. Huggins wrote:
> >
> > David, you are quite right.  There is a bug in DJGPP as compared to
> > Borland's scanf function.  I don't really know how to cure the problem
> > except for not using scanf.  I personally do not use scanf as it is too
> > limiting .  I would suggest using getch or getche depending on whether  or
> > not you want the entered chartered echoed to the screen.
> 
> It's not a bug; Borland uses a highly nonstandard fflush() that somehow
> manages to work on input streams.  According to the ANSI standard,
> flushing an input stream is an undefined operation.

Given that it's undefined, perhaps this patch would be appropriate.

*** src/libc/ansi/stdio/fflush.c~       Sun Jun 28 23:36:20 1998
--- src/libc/ansi/stdio/fflush.c        Mon Jul 13 22:48:34 1998
***************
*** 29,34 ****
--- 29,37 ----
      return 0;
    }
  
+   if (f->_flag & _IOREAD)
+     for (;;) printf("Read the ANSI standard, Borland luser!\n");
+ 
    f->_flag &= ~_IOUNGETC;
    if ((f->_flag&(_IONBF|_IOWRT))==_IOWRT
        && (base = f->_base) != NULL



:))
-- 

Nate Eldredge
nate AT cartsys DOT com