From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10109142229.AA18780@clio.rice.edu>
Subject: Re: A new W2K problem: djtar
To: djgpp-workers AT delorie DOT com
Date: Fri, 14 Sep 2001 17:29:24 -0500 (CDT)
In-Reply-To: <8296-Fri14Sep2001131608+0300-eliz@is.elta.co.il> from "Eli Zaretskii" at Sep 14, 2001 01:16:09 PM
X-Mailer: ELM [version 2.5 PL2]
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
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

> > There seems to be a new problem with W2K: djtar prints error messages
> > about EBADF when  run on a .tar.gz archive.  I'm trying to debug this.
> 
> Okay, I think I understand the problem.  Here's the relevant fragment
> from epunzip.c:
> 
>       if(!should_be_written)
> 	epoutfile = open("/dev/null",
> 			 O_WRONLY | O_BINARY | O_CREAT | O_EXCL,
> 			 S_IWRITE | S_IREAD);
> 
> This fails inside `open', since it sees O_CREAT, decides that the file
> should be created anew, calls __file_exists to see if /dev/null
> exists, and since __file_exists returns non-zero, `open' fails.
> 
> This seems like a bug in epunzip.c, doesn't it?  Why should it insist
> on creating /dev/null?
> 
> OTOH, I don't understand how does it work on other platforms?  Oh, I
> see: __file_exists fails for devices on every other platform.  Does
> this form of `open' work on GNU/Linux?

It seems we should not be trying to O_CREAT devices, but if that's
common for devices (e.g. > /dev/null) maybe we need to remove the quick 
fail logic in open and replace with something that knows about devices. 
Depending on a feature of __file_exists (_chmod) failing on devices 
seems dangerous.  I'll look at it.