Date: Fri, 23 Mar 2001 13:26:33 +0200
From: "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>
Sender: halo1 AT zahav DOT net DOT il
To: "Mark E." <snowball3 AT bigfoot DOT com>
Message-Id: <9743-Fri23Mar2001132632+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-workers AT delorie DOT com
In-reply-to: <3ABA9AB6.9970.50AA59@localhost> (snowball3@bigfoot.com)
Subject: Re: pipe() emulation
References:  <3ABA9AB6 DOT 9970 DOT 50AA59 AT localhost>
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> From: "Mark E." <snowball3 AT bigfoot DOT com>
> Date: Fri, 23 Mar 2001 00:37:10 -0500
> 
> After looking over the POSIX docs for pipe, I took out the code that moved up 
> the handles since it goes against at least two sentences in the draft 
> standard.

Which ones?

> To make the emulation more complete, I could 
> even add code to prevent lseek from working with files opened with pipe since 
> POSIX explicitly forbids that.

I think this is a good idea.  In general, if better Posix compliance
is possible without too much effort, I think we should have it.

> ! This function creates a pipe and places a file descriptor for the read end
> ! of the pipe in @var{fildes[0]}, and another for the write end in
> ! @var{fildes[1]}.  Data written to @var{fildes[1]} will be read from
> ! @var{fildes[0]} in a first-in first-out (FIFO) basis.
>   
> + Note that DOS doesn't support pipes so they are emulated with temporary
> + files.  This function should only be used in porting Unix and POSIX programs
> + that use it.

I think this should say that fork/exec and any other way to get async
subprocesses will not work with this pipe implementation.  Also, an
example, perhaps from real Bash code, would be nice.