Date: Thu, 23 Aug 2001 09:10:48 +0300
From: "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>
Sender: halo1 AT zahav DOT net DOT il
To: sandmann AT clio DOT rice DOT edu
Message-Id: <2110-Thu23Aug2001091047+0300-eliz@is.elta.co.il>
X-Mailer: Emacs 20.6 (via feedmail 8.3.emacs20_6 I) and Blat ver 1.8.9
CC: djgpp-workers AT delorie DOT com
In-reply-to: <10108221941.AA14151@clio.rice.edu> (sandmann@clio.rice.edu)
Subject: Re: Fseek on STDIN problem on Win 2K
References:  <10108221941 DOT AA14151 AT clio DOT rice DOT edu>
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: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
> Date: Wed, 22 Aug 2001 14:41:04 -0500 (CDT)
> 
> does the 2.03 library not call the is_executable?

Yes, it does.

> fstat() doesn't cause this problem in 2.03 which might explain why
> it's more rare).

So perhaps we should find out which change since v2.03 triggers this.

> I tried test >> testout
> I tried test >> nul
> 
> fstat() is OK with handle 1, in both cases thinks it's a file, lseek
> says it's at the beginning (0) wrong for file append.

lseek returning 0 on append is correct for NUL, since NUL is always
empty.  Does lseek say it's at the beginning for a non-empty file
testout as well?  If so, this doesn't happen on Windows 98.

> First example gets the right size in fstat, second example says it
> is zero length.

Same thing happens here on Windows 98, and zero length is okay with
NUL.  (I'm assuming that by ``second example'' you mean "test >>nul".)

> Both examples return times (first example is file time, second example
> is weird date in future).

The time for NUL should be the current time, but that code only works
if fstat detects that it's a device, which I presume isn't happening
on NT?

> In both cases _read fails on the handle.

Fails how?  Do you mean to say that NT disallows reads on stdout?  It
does work on Windows 98: I redirected stdout into a raw COFF file, and
fstat readily reported it's an executable.

> Proposal:  fstat() (or dev info?) does the following:
>   if dev_info == 0 and NT and handlenumb <5 dev_info = 0x8000

How is this better than leave the current problems with fstat intact?
(Admittedly, I don't remember anymore why this was a grave problem ;-)

Also, I think Bash uses handle 5 quite a lot in configure scripts, and
the programs it invokes inherit the redirection.  (Mark, am I right?)
Does redirection from Bash have this problem?  (I think you said no,
but I'm not sure.)