Date: Mon, 01 Jan 2001 20:49:06 +0200
From: "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>
Sender: halo1 AT zahav DOT net DOT il
To: jmichel AT schur DOT institut DOT math DOT jussieu DOT fr (Jean Michel)
Message-Id: <2561-Mon01Jan2001204906+0200-eliz@is.elta.co.il>
X-Mailer: Emacs 20.6 (via feedmail 8.3.emacs20_6 I) and Blat ver 1.8.6
CC: djgpp AT delorie DOT com
In-reply-to: <92qb0f$17r$1@schur.institut.math.jussieu.fr>
	(jmichel AT schur DOT institut DOT math DOT jussieu DOT fr)
Subject: Re: Bug in djgpp libc
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1010101162145 DOT 9706D-100000 AT is> <92qb0f$17r$1 AT schur DOT institut DOT math DOT jussieu DOT fr>
Reply-To: djgpp AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> From: jmichel AT schur DOT institut DOT math DOT jussieu DOT fr (Jean Michel)
> Newsgroups: comp.os.msdos.djgpp
> Date: 1 Jan 2001 17:24:47 +0100
> ...
> >The program can know that it writes to a console device: it just needs to 
> >call isatty() on the handle.  It can also filter out ^Z characters if 
> >stdout is connected to the console.  Finally, it can remove trailing ^Z 
> >from input after it reads it.  (Such removal will happen automatically if 
> >the input is read in text mode.)
> >
> >Would that solve the problem?
> 
> Well, does not djgpp  claim to be an environment which  makes it easy to
> port  Unix programs  to DOS  (even  at the  expense  of doing  a lot  of
> contortions to implement stat accurately, for instance)?

We try, yes.

> I claim that in
> the same  way it  should be the  job of the  libc low-level  routines to
> detect that  the output is to  a console device, and  do the appropriate
> thing (filter ^Z,  whatever) so that no spurious error  is returned

This was discussed on the developers' list several times in the past.
Unfortunately, the conclusion was that there's no way of fixing this
without either breaking something else or fatally hurting performance
of programs which don't send ^Z to the console (those are the
majority, btw).  For example, if low-level libc primitives would need
to scan each buffer they get for ^Z characters, instead of simply
handing it to DOS, they will become much slower.

I'd still like to know if the suggestions I made help you solve the
problem, and if not, why.

One other possibility is to use the DJGPP termios emulation, which
modifies the console I/O so that ^Z is no longer a problem.  If ruby
is an interactive program coming from Unix, it probably tries to use
termios anyway; perhaps the DJGPP port disables that for some
(histerical?)  reasons, in which case all you need to do is remove the
disabling code.