Date: Tue, 07 Aug 2001 20:19:31 +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: <7458-Tue07Aug2001201931+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: acottrel AT ihug DOT com DOT au, djgpp-workers AT delorie DOT com
In-reply-to: <10108071458.AA17574@clio.rice.edu> (sandmann@clio.rice.edu)
Subject: Re: Windows 2000 LFN issues
References:  <10108071458 DOT AA17574 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: Tue, 7 Aug 2001 09:58:03 -0500 (CDT)
> 
> > > Just to give you hope, out of the zip "touch" works with LFNs on XP RC1.
> > > So they occasionally fix some things :-)
> > 
> > Sorry, I don't understand: what is this ``out of the zip "touch"''?
> > Do you mean that you reported the bug and it already got corrected? ;-)
> 
> The touch executable doesn't work on Windows 2000 if LFN is on, but it
> does work without modifictions or rebuilding on XP.

That's good to know.

But the problems with IOCTL and in particular the "> nul" issue are
still there, yes?

> > Btw, if XP behaves differently from W2K, we will need a way to
> > distinguish between them.  Right now, it's not possible, AFAIK,
> > without some heavy and dangerous processing that is inappropriate for
> > libc.a.
> 
> Right now I haven't found anything new under XP to require different
> behavior.

Well, the fact that utime is broken on W2K but not on XP seems to be
one ;-)

> But we do have 3 environments which are very similar from
> DOS version but may require differences in code:
> 
>  1) Windows NT 4.0 with LFN TSR
>  2) Windows 2000
>  3) Windows XP (I haven't tried the personal version, just pro)
> 
> The difference between 1 and 2 worries me.

I'm worried by all three of these: they look identical to a DJGPP
program, as far as DOS version is concerned, but they don't behave in
identical ways.  We will need to keep all three configurations (and
also NT4 without LFN) in mind when we design the W2K/XP work-arounds,
and test them.

> I was also thinking about what happens if we don't fix create handles.
> the IOCTLs break - so it looks like a file on drive A (not too bad).

I intended to grep the library for IOCTL calls, then for calls to
IOCTL callers, then for calls to the callers of IOCTL callers, etc.
That's the only way to assess the possible damage.  I'm afraid that
IOCTL breakage will cause all kinds of rare and subtle bugs and
misfeatures, even though 90% of cases might just work.  I think that
we never heard much about IOCTL until now because it's used by
relatively exotic features, like getmntent, statfs, fcntl (which
didn't exist in v2.03), etc.  But when you do use those features, you
need working IOCTL.

> The utime stuff may be fixable with the direct call.

Yes, that seems the best alternative for this particular case.

There's also `findfirst' on devices, which will probably affect `stat'
and `access' as well.  We should repair that because some programs, I
forget which, try `access' on a file/device before doing I/O.  And
`stat' is of course called by lots of Unix-born programs...