Date: Wed, 13 Jan 1999 15:56:36 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: "Gary V. Vaughan" <gvaughan AT oranda DOT demon DOT co DOT uk>
cc: Alexandre Oliva <oliva AT dcc DOT unicamp DOT br>, Andris Pavenis <pavenis AT lanet DOT lv>,
        "Mark E." <snowball3 AT usa DOT net>, autoconf AT gnu DOT org,
        djgpp-workers AT delorie DOT com
Subject: Re: patch for autoheader.sh
In-Reply-To: <369C7BB6.DBB43A78@oranda.demon.co.uk>
Message-ID: <Pine.SUN.3.91.990113155236.14951E-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: djgpp-workers AT delorie DOT com


On Wed, 13 Jan 1999, Gary V. Vaughan wrote:

> > Ugh!  Better avoid it.  /dev/null is so abused that we'd die before
> > we could replace all uses of it with the result of the configure
> > test.
> 
> I don't want to die.  Lets not mess with configure then.  Perhaps the
> win95 failure modes are pathological... and we can ignore the problem
> for the general case?

Sorry, you cannot ignore this.  Neither MS-DOS nor Windows 9X can handle 
a file being renamed under their feet while it is open.  DOS just goes 
amok (how much amok depends on how large the file being renamed is, and 
what optional software, like a disk cache, is installed); Windows 9X 
simply fails the call.  Only Windows NT handles these cases like Unix 
does.

If /dev/null is verboten, maybe what Andris suggested with 1>&2 is 
better?